Mr. Somesh Kumar Dewangan, Disha Institute of Management and Technology, India .... Dr. Ekta Walia Bhullar, Maharishi Markandeshwar University, Mullana Ambala (Haryana), ...... Furthermore, m aintaining q ueue stab ility is important as.
IJCSI
International Journal of Computer Science Issues
Volume 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
© IJCSI PUBLICATION www.IJCSI.org
IJCSI proceedings are currently indexed by:
© IJCSI PUBLICATION 2010 www.IJCSI.org
IJCSI Publicity Board 2010 Dr. Borislav D Dimitrov Department of General Practice, Royal College of Surgeons in Ireland Dublin, Ireland
Dr. Vishal Goyal Department of Computer Science, Punjabi University Patiala, India
Mr. Nehinbe Joshua University of Essex Colchester, Essex, UK
Mr. Vassilis Papataxiarhis Department of Informatics and Telecommunications National and Kapodistrian University of Athens, Athens, Greece
EDITORIAL In this third edition of 2010, we bring forward issues from va rious dynam ic co mputer science areas ranging from system perf ormance, computer vision, artifi cial intelligence, software engineering, m ultimedia , pattern recognition, info rmation retrieval, d atabases, security and networking among others. As always we thank all our review ers for prov iding constructive comm ents on papers sent to them for review. This helps enorm ously in improving the quality of papers published in this issue. IJCSI will maintain its policy of sending print copies of the journal to all co rresponding authors worldwide free of charg e. Apart from availability of the f ull-texts from the jou rnal website, all published p apers are d eposited in open-access repositories to m ake access eas ier and ensur e continuous availability of its proceedings. The trans ition from the 2 nd issu e to the 3 rd on e has been m arked with an agreem ent signed between IJCSI and ProQuest an d EBSCOHOST, two lead ing directo ries to help in th e dissemination of our published papers. W e believe further indexing and more dissem ination will definitely lead to further citations of our authors’ articles. We are pleased to present IJCSI Volum e 7, Issu e 3, May 2010, split in el even numbers (IJCSI Vol. 7, Issue 3, No. 5). The acceptance rate for this issue is 37.88%. We wish you a happy reading! IJCSI Editorial Board May 2010 Issue ISSN (Print): 1694-0814 ISSN (Online): 1694-0784 © IJCSI Publications www.IJCSI.org
IJCSI Editorial Board 2010 Dr Tristan Vanrullen Chief Editor LPL, Laboratoire Parole et Langage - CNRS - Aix en Provence, France LABRI, Laboratoire Bordelais de Recherche en Informatique - INRIA - Bordeaux, France LEEE, Laboratoire d'Esthétique et Expérimentations de l'Espace - Université d'Auvergne, France Dr Constantino Malagôn Associate Professor Nebrija University Spain Dr Lamia Fourati Chaari Associate Professor Multimedia and Informatics Higher Institute in SFAX Tunisia Dr Mokhtar Beldjehem Professor Sainte-Anne University Halifax, NS, Canada Dr Pascal Chatonnay Assistant Professor MaÎtre de Conférences Laboratoire d'Informatique de l'Université de Franche-Comté Université de Franche-Comté France Dr Yee-Ming Chen Professor Department of Industrial Engineering and Management Yuan Ze University Taiwan
Dr Vishal Goyal Assistant Professor Department of Computer Science Punjabi University Patiala, India Dr Natarajan Meghanathan Assistant Professor REU Program Director Department of Computer Science Jackson State University Jackson, USA Dr Deepak Laxmi Narasimha Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia Dr Navneet Agrawal Assistant Professor Department of ECE, College of Technology & Engineering, MPUAT, Udaipur 313001 Rajasthan, India Prof N. Jaisankar Assistant Professor School of Computing Sciences, VIT University Vellore, Tamilnadu, India
IJCSI Reviewers Committee 2010 Mr. Markus Schatten, University of Zagreb, Faculty of Organization and Informatics, Croatia Mr. Vassilis Papataxiarhis, Department of Informatics and Telecommunications, National and Kapodistrian University of Athens, Athens, Greece Dr Modestos Stavrakis, University of the Aegean, Greece Dr Fadi KHALIL, LAAS -- CNRS Laboratory, France Dr Dimitar Trajanov, Faculty of Electrical Engineering and Information technologies, ss. Cyril and Methodius Univesity - Skopje, Macedonia Dr Jinping Yuan, College of Information System and Management,National Univ. of Defense Tech., China Dr Alexis Lazanas, Ministry of Education, Greece Dr Stavroula Mougiakakou, University of Bern, ARTORG Center for Biomedical Engineering Research, Switzerland Dr Cyril de Runz, CReSTIC-SIC, IUT de Reims, University of Reims, France Mr. Pramodkumar P. Gupta, Dept of Bioinformatics, Dr D Y Patil University, India Dr Alireza Fereidunian, School of ECE, University of Tehran, Iran Mr. Fred Viezens, Otto-Von-Guericke-University Magdeburg, Germany Dr. Richard G. Bush, Lawrence Technological University, United States Dr. Ola Osunkoya, Information Security Architect, USA Mr. Kotsokostas N.Antonios, TEI Piraeus, Hellas Prof Steven Totosy de Zepetnek, U of Halle-Wittenberg & Purdue U & National Sun Yat-sen U, Germany, USA, Taiwan Mr. M Arif Siddiqui, Najran University, Saudi Arabia Ms. Ilknur Icke, The Graduate Center, City University of New York, USA Prof Miroslav Baca, Faculty of Organization and Informatics, University of Zagreb, Croatia Dr. Elvia Ruiz Beltrán, Instituto Tecnológico de Aguascalientes, Mexico Mr. Moustafa Banbouk, Engineer du Telecom, UAE Mr. Kevin P. Monaghan, Wayne State University, Detroit, Michigan, USA Ms. Moira Stephens, University of Sydney, Australia Ms. Maryam Feily, National Advanced IPv6 Centre of Excellence (NAV6) , Universiti Sains Malaysia (USM), Malaysia Dr. Constantine YIALOURIS, Informatics Laboratory Agricultural University of Athens, Greece Mrs. Angeles Abella, U. de Montreal, Canada Dr. Patrizio Arrigo, CNR ISMAC, italy Mr. Anirban Mukhopadhyay, B.P.Poddar Institute of Management & Technology, India Mr. Dinesh Kumar, DAV Institute of Engineering & Technology, India Mr. Jorge L. Hernandez-Ardieta, INDRA SISTEMAS / University Carlos III of Madrid, Spain Mr. AliReza Shahrestani, University of Malaya (UM), National Advanced IPv6 Centre of Excellence (NAv6), Malaysia Mr. Blagoj Ristevski, Faculty of Administration and Information Systems Management - Bitola, Republic of Macedonia Mr. Mauricio Egidio Cantão, Department of Computer Science / University of São Paulo, Brazil Mr. Jules Ruis, Fractal Consultancy, The Netherlands
Mr. Mohammad Iftekhar Husain, University at Buffalo, USA Dr. Deepak Laxmi Narasimha, Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, Malaysia Dr. Paola Di Maio, DMEM University of Strathclyde, UK Dr. Bhanu Pratap Singh, Institute of Instrumentation Engineering, Kurukshetra University Kurukshetra, India Mr. Sana Ullah, Inha University, South Korea Mr. Cornelis Pieter Pieters, Condast, The Netherlands Dr. Amogh Kavimandan, The MathWorks Inc., USA Dr. Zhinan Zhou, Samsung Telecommunications America, USA Mr. Alberto de Santos Sierra, Universidad Politécnica de Madrid, Spain Dr. Md. Atiqur Rahman Ahad, Department of Applied Physics, Electronics & Communication Engineering (APECE), University of Dhaka, Bangladesh Dr. Charalampos Bratsas, Lab of Medical Informatics, Medical Faculty, Aristotle University, Thessaloniki, Greece Ms. Alexia Dini Kounoudes, Cyprus University of Technology, Cyprus Mr. Anthony Gesase, University of Dar es salaam Computing Centre, Tanzania Dr. Jorge A. Ruiz-Vanoye, Universidad Juárez Autónoma de Tabasco, Mexico Dr. Alejandro Fuentes Penna, Universidad Popular Autónoma del Estado de Puebla, México Dr. Ocotlán Díaz-Parra, Universidad Juárez Autónoma de Tabasco, México Mrs. Nantia Iakovidou, Aristotle University of Thessaloniki, Greece Mr. Vinay Chopra, DAV Institute of Engineering & Technology, Jalandhar Ms. Carmen Lastres, Universidad Politécnica de Madrid - Centre for Smart Environments, Spain Dr. Sanja Lazarova-Molnar, United Arab Emirates University, UAE Mr. Srikrishna Nudurumati, Imaging & Printing Group R&D Hub, Hewlett-Packard, India Dr. Olivier Nocent, CReSTIC/SIC, University of Reims, France Mr. Burak Cizmeci, Isik University, Turkey Dr. Carlos Jaime Barrios Hernandez, LIG (Laboratory Of Informatics of Grenoble), France Mr. Md. Rabiul Islam, Rajshahi university of Engineering & Technology (RUET), Bangladesh Dr. LAKHOUA Mohamed Najeh, ISSAT - Laboratory of Analysis and Control of Systems, Tunisia Dr. Alessandro Lavacchi, Department of Chemistry - University of Firenze, Italy Mr. Mungwe, University of Oldenburg, Germany Mr. Somnath Tagore, Dr D Y Patil University, India Ms. Xueqin Wang, ATCS, USA Dr. Borislav D Dimitrov, Department of General Practice, Royal College of Surgeons in Ireland, Dublin, Ireland Dr. Fondjo Fotou Franklin, Langston University, USA Dr. Vishal Goyal, Department of Computer Science, Punjabi University, Patiala, India Mr. Thomas J. Clancy, ACM, United States Dr. Ahmed Nabih Zaki Rashed, Dr. in Electronic Engineering, Faculty of Electronic Engineering, menouf 32951, Electronics and Electrical Communication Engineering Department, Menoufia university, EGYPT, EGYPT Dr. Rushed Kanawati, LIPN, France Mr. Koteshwar Rao, K G Reddy College Of ENGG.&TECH,CHILKUR, RR DIST.,AP, India
Mr. M. Nagesh Kumar, Department of Electronics and Communication, J.S.S. research foundation, Mysore University, Mysore-6, India Dr. Ibrahim Noha, Grenoble Informatics Laboratory, France Mr. Muhammad Yasir Qadri, University of Essex, UK Mr. Annadurai .P, KMCPGS, Lawspet, Pondicherry, India, (Aff. Pondicherry Univeristy, India Mr. E Munivel , CEDTI (Govt. of India), India Dr. Chitra Ganesh Desai, University of Pune, India Mr. Syed, Analytical Services & Materials, Inc., USA Dr. Mashud Kabir, Department of Computer Science, University of Tuebingen, Germany Mrs. Payal N. Raj, Veer South Gujarat University, India Mrs. Priti Maheshwary, Maulana Azad National Institute of Technology, Bhopal, India Mr. Mahesh Goyani, S.P. University, India, India Mr. Vinay Verma, Defence Avionics Research Establishment, DRDO, India Dr. George A. Papakostas, Democritus University of Thrace, Greece Mr. Abhijit Sanjiv Kulkarni, DARE, DRDO, India Mr. Kavi Kumar Khedo, University of Mauritius, Mauritius Dr. B. Sivaselvan, Indian Institute of Information Technology, Design & Manufacturing, Kancheepuram, IIT Madras Campus, India Dr. Partha Pratim Bhattacharya, Greater Kolkata College of Engineering and Management, West Bengal University of Technology, India Mr. Manish Maheshwari, Makhanlal C University of Journalism & Communication, India Dr. Siddhartha Kumar Khaitan, Iowa State University, USA Dr. Mandhapati Raju, General Motors Inc, USA Dr. M.Iqbal Saripan, Universiti Putra Malaysia, Malaysia Mr. Ahmad Shukri Mohd Noor, University Malaysia Terengganu, Malaysia Mr. Selvakuberan K, TATA Consultancy Services, India Dr. Smita Rajpal, Institute of Technology and Management, Gurgaon, India Mr. Rakesh Kachroo, Tata Consultancy Services, India Mr. Raman Kumar, National Institute of Technology, Jalandhar, Punjab., India Mr. Nitesh Sureja, S.P.University, India Dr. M. Emre Celebi, Louisiana State University, Shreveport, USA Dr. Aung Kyaw Oo, Defence Services Academy, Myanmar Mr. Sanjay P. Patel, Sankalchand Patel College of Engineering, Visnagar, Gujarat, India Dr. Pascal Fallavollita, Queens University, Canada Mr. Jitendra Agrawal, Rajiv Gandhi Technological University, Bhopal, MP, India Mr. Ismael Rafael Ponce Medellín, Cenidet (Centro Nacional de Investigación y Desarrollo Tecnológico), Mexico Mr. Supheakmungkol SARIN, Waseda University, Japan Mr. Shoukat Ullah, Govt. Post Graduate College Bannu, Pakistan Dr. Vivian Augustine, Telecom Zimbabwe, Zimbabwe Mrs. Mutalli Vatila, Offshore Business Philipines, Philipines Dr. Emanuele Goldoni, University of Pavia, Dept. of Electronics, TLC & Networking Lab, Italy Mr. Pankaj Kumar, SAMA, India Dr. Himanshu Aggarwal, Punjabi University,Patiala, India Dr. Vauvert Guillaume, Europages, France
Prof Yee Ming Chen, Department of Industrial Engineering and Management, Yuan Ze University, Taiwan Dr. Constantino Malagón, Nebrija University, Spain Prof Kanwalvir Singh Dhindsa, B.B.S.B.Engg.College, Fatehgarh Sahib (Punjab), India Mr. Angkoon Phinyomark, Prince of Singkla University, Thailand Ms. Nital H. Mistry, Veer Narmad South Gujarat University, Surat, India Dr. M.R.Sumalatha, Anna University, India Mr. Somesh Kumar Dewangan, Disha Institute of Management and Technology, India Mr. Raman Maini, Punjabi University, Patiala(Punjab)-147002, India Dr. Abdelkader Outtagarts, Alcatel-Lucent Bell-Labs, France Prof Dr. Abdul Wahid, AKG Engg. College, Ghaziabad, India Mr. Prabu Mohandas, Anna University/Adhiyamaan College of Engineering, india Dr. Manish Kumar Jindal, Panjab University Regional Centre, Muktsar, India Prof Mydhili K Nair, M S Ramaiah Institute of Technnology, Bangalore, India Dr. C. Suresh Gnana Dhas, VelTech MultiTech Dr.Rangarajan Dr.Sagunthala Engineering College,Chennai,Tamilnadu, India Prof Akash Rajak, Krishna Institute of Engineering and Technology, Ghaziabad, India Mr. Ajay Kumar Shrivastava, Krishna Institute of Engineering & Technology, Ghaziabad, India Mr. Deo Prakash, SMVD University, Kakryal(J&K), India Dr. Vu Thanh Nguyen, University of Information Technology HoChiMinh City, VietNam Prof Deo Prakash, SMVD University (A Technical University open on I.I.T. Pattern) Kakryal (J&K), India Dr. Navneet Agrawal, Dept. of ECE, College of Technology & Engineering, MPUAT, Udaipur 313001 Rajasthan, India Mr. Sufal Das, Sikkim Manipal Institute of Technology, India Mr. Anil Kumar, Sikkim Manipal Institute of Technology, India Dr. B. Prasanalakshmi, King Saud University, Saudi Arabia. Dr. K D Verma, S.V. (P.G.) College, Aligarh, India Mr. Mohd Nazri Ismail, System and Networking Department, University of Kuala Lumpur (UniKL), Malaysia Dr. Nguyen Tuan Dang, University of Information Technology, Vietnam National University Ho Chi Minh city, Vietnam Dr. Abdul Aziz, University of Central Punjab, Pakistan Dr. P. Vasudeva Reddy, Andhra University, India Mrs. Savvas A. Chatzichristofis, Democritus University of Thrace, Greece Mr. Marcio Dorn, Federal University of Rio Grande do Sul - UFRGS Institute of Informatics, Brazil Mr. Luca Mazzola, University of Lugano, Switzerland Mr. Nadeem Mahmood, Department of Computer Science, University of Karachi, Pakistan Mr. Hafeez Ullah Amin, Kohat University of Science & Technology, Pakistan Dr. Professor Vikram Singh, Ch. Devi Lal University, Sirsa (Haryana), India Mr. M. Azath, Calicut/Mets School of Enginerring, India Dr. J. Hanumanthappa, DoS in CS, University of Mysore, India Dr. Shahanawaj Ahamad, Department of Computer Science, King Saud University, Saudi Arabia Dr. K. Duraiswamy, K. S. Rangasamy College of Technology, India Prof. Dr Mazlina Esa, Universiti Teknologi Malaysia, Malaysia
Dr. P. Vasant, Power Control Optimization (Global), Malaysia Dr. Taner Tuncer, Firat University, Turkey Dr. Norrozila Sulaiman, University Malaysia Pahang, Malaysia Prof. S K Gupta, BCET, Guradspur, India Dr. Latha Parameswaran, Amrita Vishwa Vidyapeetham, India Mr. M. Azath, Anna University, India Dr. P. Suresh Varma, Adikavi Nannaya University, India Prof. V. N. Kamalesh, JSS Academy of Technical Education, India Dr. D Gunaseelan, Ibri College of Technology, Oman Mr. Sanjay Kumar Anand, CDAC, India Mr. Akshat Verma, CDAC, India Mrs. Fazeela Tunnisa, Najran University, Kingdom of Saudi Arabia Mr. Hasan Asil, Islamic Azad University Tabriz Branch (Azarshahr), Iran Prof. Dr Sajal Kabiraj, Fr. C Rodrigues Institute of Management Studies (Affiliated to University of Mumbai, India), India Mr. Syed Fawad Mustafa, GAC Center, Shandong University, China Dr. Natarajan Meghanathan, Jackson State University, Jackson, MS, USA Prof. Selvakani Kandeeban, Francis Xavier Engineering College, India Mr. Tohid Sedghi, Urmia University, Iran Dr. S. Sasikumar, PSNA College of Engg and Tech, Dindigul, India Dr. Anupam Shukla, Indian Institute of Information Technology and Management Gwalior, India Mr. Rahul Kala, Indian Institute of Inforamtion Technology and Management Gwalior, India Dr. A V Nikolov, National University of Lesotho, Lesotho Mr. Kamal Sarkar, Department of Computer Science and Engineering, Jadavpur University, India Dr. Mokhled S. AlTarawneh, Computer Engineering Dept., Faculty of Engineering, Mutah University, Jordan, Jordan Prof. Sattar J Aboud, Iraqi Council of Representatives, Iraq-Baghdad Dr. Prasant Kumar Pattnaik, Department of CSE, KIST, India Dr. Mohammed Amoon, King Saud University, Saudi Arabia Dr. Tsvetanka Georgieva, Department of Information Technologies, St. Cyril and St. Methodius University of Veliko Tarnovo, Bulgaria Dr. Eva Volna, University of Ostrava, Czech Republic Mr. Ujjal Marjit, University of Kalyani, West-Bengal, India Dr. Prasant Kumar Pattnaik, KIST,Bhubaneswar,India, India Dr. Guezouri Mustapha, Department of Electronics, Faculty of Electrical Engineering, University of Science and Technology (USTO), Oran, Algeria Mr. Maniyar Shiraz Ahmed, Najran University, Najran, Saudi Arabia Dr. Sreedhar Reddy, JNTU, SSIETW, Hyderabad, India Mr. Bala Dhandayuthapani Veerasamy, Mekelle University, Ethiopa Mr. Arash Habibi Lashkari, University of Malaya (UM), Malaysia Mr. Rajesh Prasad, LDC Institute of Technical Studies, Allahabad, India Ms. Habib Izadkhah, Tabriz University, Iran Dr. Lokesh Kumar Sharma, Chhattisgarh Swami Vivekanand Technical University Bhilai, India Mr. Kuldeep Yadav, IIIT Delhi, India Dr. Naoufel Kraiem, Institut Superieur d'Informatique, Tunisia
Prof. Frank Ortmeier, Otto-von-Guericke-Universitaet Magdeburg, Germany Mr. Ashraf Aljammal, USM, Malaysia Mrs. Amandeep Kaur, Department of Computer Science, Punjabi University, Patiala, Punjab, India Mr. Babak Basharirad, University Technology of Malaysia, Malaysia Mr. Avinash singh, Kiet Ghaziabad, India Dr. Miguel Vargas-Lombardo, Technological University of Panama, Panama Dr. Tuncay Sevindik, Firat University, Turkey Ms. Pavai Kandavelu, Anna University Chennai, India Mr. Ravish Khichar, Global Institute of Technology, India Mr Aos Alaa Zaidan Ansaef, Multimedia University, Cyberjaya, Malaysia Dr. Awadhesh Kumar Sharma, Dept. of CSE, MMM Engg College, Gorakhpur-273010, UP, India Mr. Qasim Siddique, FUIEMS, Pakistan Dr. Le Hoang Thai, University of Science, Vietnam National University - Ho Chi Minh City, Vietnam Dr. Saravanan C, NIT, Durgapur, India Dr. Vijay Kumar Mago, DAV College, Jalandhar, India Dr. Do Van Nhon, University of Information Technology, Vietnam Mr. Georgios Kioumourtzis, University of Patras, Greece Mr. Amol D.Potgantwar, SITRC Nasik, India Mr. Lesedi Melton Masisi, Council for Scientific and Industrial Research, South Africa Dr. Karthik.S, Department of Computer Science & Engineering, SNS College of Technology, India Mr. Nafiz Imtiaz Bin Hamid, Department of Electrical and Electronic Engineering, Islamic University of Technology (IUT), Bangladesh Mr. Muhammad Imran Khan, Universiti Teknologi PETRONAS, Malaysia Dr. Abdul Kareem M. Radhi, Information Engineering - Nahrin University, Iraq Dr. Mohd Nazri Ismail, University of Kuala Lumpur, Malaysia Dr. Manuj Darbari, BBDNITM, Institute of Technology, A-649, Indira Nagar, Lucknow 226016, India Ms. Izerrouken, INP-IRIT, France Mr. Nitin Ashokrao Naik, Dept. of Computer Science, Yeshwant Mahavidyalaya, Nanded, India Mr. Nikhil Raj, National Institute of Technology, Kurukshetra, India Prof. Maher Ben Jemaa, National School of Engineers of Sfax, Tunisia Prof. Rajeshwar Singh, BRCM College of Engineering and Technology, Bahal Bhiwani, Haryana, India Mr. Gaurav Kumar, Department of Computer Applications, Chitkara Institute of Engineering and Technology, Rajpura, Punjab, India Mr. Ajeet Kumar Pandey, Indian Institute of Technology, Kharagpur, India Mr. Rajiv Phougat, IBM Corporation, USA Mrs. Aysha V, College of Applied Science Pattuvam affiliated with Kannur University, India Dr. Debotosh Bhattacharjee, Department of Computer Science and Engineering, Jadavpur University, Kolkata-700032, India Dr. Neelam Srivastava, Institute of engineering & Technology, Lucknow, India Prof. Sweta Verma, Galgotia's College of Engineering & Technology, Greater Noida, India Mr. Harminder Singh BIndra, MIMIT, INDIA Dr. Lokesh Kumar Sharma, Chhattisgarh Swami Vivekanand Technical University, Bhilai, India Mr. Tarun Kumar, U.P. Technical University/Radha Govinend Engg. College, India Mr. Tirthraj Rai, Jawahar Lal Nehru University, New Delhi, India
Mr. Akhilesh Tiwari, Madhav Institute of Technology & Science, India Mr. Dakshina Ranjan Kisku, Dr. B. C. Roy Engineering College, WBUT, India Ms. Anu Suneja, Maharshi Markandeshwar University, Mullana, Haryana, India Mr. Munish Kumar Jindal, Punjabi University Regional Centre, Jaito (Faridkot), India Dr. Ashraf Bany Mohammed, Management Information Systems Department, Faculty of Administrative and Financial Sciences, Petra University, Jordan Mrs. Jyoti Jain, R.G.P.V. Bhopal, India Dr. Lamia Chaari, SFAX University, Tunisia Mr. Akhter Raza Syed, Department of Computer Science, University of Karachi, Pakistan Prof. Khubaib Ahmed Qureshi, Information Technology Department, HIMS, Hamdard University, Pakistan Prof. Boubker Sbihi, Ecole des Sciences de L'Information, Morocco Dr. S. M. Riazul Islam, Inha University, South Korea Prof. Lokhande S.N., S.R.T.M.University, Nanded (MH), India Dr. Vijay H Mankar, Dept. of Electronics, Govt. Polytechnic, Nagpur, India Dr. M. Sreedhar Reddy, JNTU, Hyderabad, SSIETW, India Mr. Ojesanmi Olusegun, Ajayi Crowther University, Oyo, Nigeria Ms. Mamta Juneja, RBIEBT, PTU, India Dr. Ekta Walia Bhullar, Maharishi Markandeshwar University, Mullana Ambala (Haryana), India Prof. Chandra Mohan, John Bosco Engineering College, India Mr. Nitin A. Naik, Yeshwant Mahavidyalaya, Nanded, India Mr. Sunil Kashibarao Nayak, Bahirji Smarak Mahavidyalaya, Basmathnagar Dist-Hingoli., India Prof. Rakesh.L, Vijetha Institute of Technology, Bangalore, India Mr B. M. Patil, Indian Institute of Technology, Roorkee, Uttarakhand, India Mr. Thipendra Pal Singh, Sharda University, K.P. III, Greater Noida, Uttar Pradesh, India Prof. Chandra Mohan, John Bosco Engg College, India Mr. Hadi Saboohi, University of Malaya - Faculty of Computer Science and Information Technology, Malaysia Dr. R. Baskaran, Anna University, India Dr. Wichian Sittiprapaporn, Mahasarakham University College of Music, Thailand Mr. Lai Khin Wee, Universiti Teknologi Malaysia, Malaysia Dr. Kamaljit I. Lakhtaria, Atmiya Institute of Technology, India Mrs. Inderpreet Kaur, PTU, Jalandhar, India Mr. Iqbaldeep Kaur, PTU / RBIEBT, India Mrs. Vasudha Bahl, Maharaja Agrasen Institute of Technology, Delhi, India Prof. Vinay Uttamrao Kale, P.R.M. Institute of Technology & Research, Badnera, Amravati, Maharashtra, India Mr. Suhas J Manangi, Microsoft, India Ms. Anna Kuzio, Adam Mickiewicz University, School of English, Poland Dr. Debojyoti Mitra, Sir Padampat Singhania University, India Prof. Rachit Garg, Department of Computer Science, L K College, India Mrs. Manjula K A, Kannur University, India Mr. Rakesh Kumar, Indian Institute of Technology Roorkee, India
TABLE OF CONTENTS
1. Survey: Grid Computing and Semantic Web Belén Bonilla-Morales Xavier Medianero-Pasco Miguel Vargas-Lombardo
Pg 1-5
2. An Enterprise Architecture Framework for Mobile Commerce Khawar Hameed Hanifa Shah Kamran Ahsan Weijun Yang
Pg 6-12
3. Performance Comparison between Active and Passive Queue Management Arash Dana Ahmad Malekloo
Pg 13-17
4. A Minimum-Process Coordinated Checkpointing Protocol For Mobile Distributed System Praveen Kumar Ajay Khunteta
Pg 23-32
5. Reputation Based Two Way Trust Model for Reliable Transactions in Grid Computing Srivaramangai P Renagaramanujam Srinivasan
Pg 33-39
6. Soft-Checkpointing Based Coordinated Checkpointing Protocol for Mobile Distributed Systems Parveen Kumar Rachit Garg
Pg 40-46
7. Footprint Recognition using Modified Sequential Haar Energy Transform (MSHET) V. D. Ambeth Kumar M. Ramakrishnan
Pg 47-51
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
1
Survey: Grid Computing and Semantic Web Belén Bonilla-Morales1, Xavier Medianero-Pasco2 and Miguel Vargas-Lombardo3 1, 2, 3
Technological University of Panama Panama City, Panama
Abstract
Grid Com puting is a p latform f or coord inated r esource sharing and problem solving on a global scale among vir tual organizations. Grid uses Grid Services to access and use a set of Grid resources. Subsequently, these Grid Ser vices need to be discov ered, selected and invoked quickly and efficiently to s atisfy the needs of a demanding environment such as Grid. In this pap er, we introd uce h ow the Grid C omputing tak es advantage of th e b enefits of Se mantic Web to manipulate Grid Services obtaining better searches, results and performance. Keywords: Grid Computing, Web Services, Grid Services, Semantic Web.
1. Introduction The Gr id is defined a s a s ecure an d co ordinated r esource sharing (direct access t o c omputers, s oftware, data, a nd high performance equipment) among groups of individuals, institutions and resources that h ave rules wh ich define what is shared, who is allowed to share resources and how resources ca n be s hared. These g roups o f i ndividuals, institutions an d resources are org anizationally an d geographically di spersed, a nd t hey a re called vi rtual organizations (VO) [1]. The benefit of i ncreasing t he c omputational p ower by combining geographically di verse sy stems and hi gh performance computational equipment raises the issues of heterogeneity and scalability [2]. The use of a middleware, such as Globus Toolkit [2, 3] , a ddresses het erogeneity, providing uniform ity throug h standa rd interfaces t o shared resources. Through a m iddleware suc h as Gl obus Toolkit, we can obtain a set of serv ices and software lib raries to support the resources of Grids and Grid applications [2]. Because Grid com puting use s Internet to e nable resource sharing of geographically diverse computational resources, we ca n use Web Services (WS) t o e nsure the communication between Grid applications.
But WS do have certain limitations and they would not be sufficient for building and communicate Grid applications. To solve th ese li mitations Grid Serv ices (GS) [2 ] are introduced. Similarly to WS [4], GS need to apply resource discovery to b e fun ctional. Ho wever, sometimes th e con ventional search of serv ices do es not pr ovide goo d r esults, in th ese cases, Sem antic Web tries t o so lve th e pro blem ad ding meaning to GS and thus, the search is improved and better results are obtained [5]. The Sem antic Web is com posed of a set of technologies , and it can be defined as a sy mbiosis of Web technologies and kno wledge rep resentation [6 ] con stituting wh at i s known as web of d ata wh ere so me h uman intelligence is integrated to t he Web, making the search easier and m ore productive [7]. Through this p aper, we will focus on demonstrating the importance o f Semantic Web f or Gri d Computing, specifically for GS. The rest of th is p aper is organ ized as follows: Section II describes t he concept of WS, t heir c haracteristics an d limitations wh en using Grid co mputing. Section III examines the concept of GS. Section IV explains the convergence b etween WS and GS; Section V desc ribes Semantic W eb t echnology and s ection IV i ntroduces semantic GS.
2. Web Services Web Services describe a di stributed c omputing pa radigm focused on In ternet-based stan dards to facilitate heterogeneous distributed computing [2]. A WS is a way to expo se th e fun ctionality o f an information system or application, and to make it available via web technologies [8]. WS a re c omposed of so ftware components or computational fu nctions w hich ha ve a set of operations
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org which define a sp ecific functionality; moreover, they have methods to allow access to the components or functions of any ser vice pr ovider from any com puter, and discovery methods t hat permit us t o i dentify a s pecific se rvice t hat the operations we need. However, WS a re m ore t han a remote proce dure invocation; WS allo w separating interface fro m implementation, this means that the interface is written in a p rogramming langu age neu tral ind ependently of t he programming language a nd pl atform t hat wa s used t o create them . For this reason, WS allow the communications among heterogeneous systems. WS are defined as a set of st andards created by the World Wide Web Con sortium (W3C) su ch as the Sim ple Obj ect Access Pr otocol (SOAP), W eb Ser vices description Language (WSDL), and Universal Description, Discovery, and Integration (UDDI) [2]. SOAP is a XM L base d com munication pr otocol that permits to exchange XML (Extensible Markup Language) messages am ong c onsumers. WSDL describes t he functionality an d in terface of WS; but it is a XML document t hat b rings t he WS i nformation [ 2, 9]. T he interface of a WS is usually called portType; a normal WS can onl y ha ve a po rtType. UDDI pr otocol is another standard of WS t hat al lows t o re gister a nd m ake Web Services available to consumers [9]. In a typical WS invocation an d discovery, t he fi rst st ep consists in discovering a WS that meets our requirem ents. Then, the UDDI registry will reply telling us what servers can provide us the WS we require. At this stage, we know the location of the WS we n eed, but we do not know how to invoke it, so we need the description of the Web Service. The WS d escribed itself u sing WSDL. Th is inform ation indicates to us the location of the WS and how to invoke it. The inv ocation is don e t hrough th e SOAP lan guage; sending a SOAP request we can ask the Web Service what we need, and th e WS an swers our question w ith a SO AP response [8]. There are some li mitations to th e use of WS in Grid Computing. WS are state less [10]. T his m eans tha t WS ca nnot remember what the end user has done from one invocation to an other. I f the en d use r wants t o perform a chai n of operations, he h as to get the resu lt of one op eration and send it as a parameter to the next operation. Furthermore, WS are persistent [10] meaning they outlive all their clients. They are a vailable from the m oment the server starts un til th e serv er sto ps. It can be po tentially insecure b ecause wh en a clien t is using a WS, all the
2
information the WS is remembering ca n be accesse d by the next clien t using t his WS. In fact, while a clien t is using a WS, another client c ould access t he same WS and ruin the operations of the first client. Due to these obstacles, a new concept arises: Grid Service [2]. GS ar e a k ind of WS th at em erge to address t he limitations of working with WS within Grid computing.
3. Grid Services Grid Se rvices are softwa re components wh ich p rovide access to a set of Grid resources such as data sources, high performance equipment and computational resources [11]. GS a re t he base of t he Open Grid Se rvices Arc hitecture (OGSA) [12], which aims to standardize all serv ices on a Grid application. OGSA defines the GS concept based on technologies o f Gri d computing and WS . It also defi nes standard mechanisms for naming, creating and discovering GS [13, 14]. Open Grid Services Infrastructure (OGSI) defines a set of conventions an d sp ecifications fo r th e use of WSDL and XML Schema to enable GS [12]. GS solve the problems WS have for their use with in Grid Computing (stateless an d pe rsistence) usi ng a factory/instance ap proach t o WS. It m eans that we c ould have a cen tral serv ice factory wh ich m aintains a set of service insta nces instead of ha ving o ne bi g st ateless service shared by all users . When a client needs to invoke an operation of th e serv ice, it wou ld talk to the in stance, not to the factory. In the same way, if th e clien t wan ts to create or destroy an in stance, it will talk to th e fact ory. Each client could have access to more than one instance of the se rvice, a nd one i nstance c ould be s hared by m any clients. T hese instances are transients (the op posed t o persistent) because th ey have a li mited lifeti me wh ich i s not bound to the lifetime of the GS container. It means that one client can create and destroy instances of a GS instead of having o ne persistent ser vice perm anently avai lable. It avoids t he risk t hat an other cl ient m ight dest roy t he operations performed over the GS by the first client. Thus, the main improvements that GS introduce are: GS support portType extension: It means we can define a p ortType a s an ext ension of a previously e xisting portType [2]. GS are state ful services: It means they retain the state after each invocation and subsequent execution [14].
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org GS are tr ansients: It m eans t hey a re not l inked t o t he lifetime of t he serve r, but they can be create d a nd destroyed as required [14]. GS have Service Data: Service Data allo ws to in clude a set of structured data to any service, which can be accessed through its int erface. Se rvice Data is an extension that allows i ncluding not only op erations in t he WSDL, but also attributes and any type of data (classe s, arrays, etc.) [8].
4. Convergence between Web Services and Grid Services Even t hough, th e go al was th e conv ergence of t he standards fo r WS (W SDL) an d GS ( OGSI), th is convergence has not achieved. The main drawbacks of OGSI are as fo llows: its lon g and dense sp ecifications, it is difficult to in tegrate it with existing WS tools, and its excessively object oriented [12]. To ov ercome th ese drawbacks and enc ourage a true convergence between the t wo types of se rvices, a ne w standard was created to replace OGSI. In contrast to OGSI, this new st andard called Web Se rvices Res ource Framework (WSRF) is based on Web Services [12]. Thus, OGSA would be base d di rectly on WS instead of being based on an improved version of OGSI Grid services, and WSRF w ould be part of a collection of W S standards rather than a patch. WSRF is not a signi ficant change but an evoluti on of OGSI. T he OGSA standa rd is not affected beca use changes affec t the infrast ructure o nly an d not the architecture [12]. WSRF provides a set of operations that WS may implement to become stateful. Applications
Applications
OGSA
OGSA
OGSI
Web Services
Web Services
3
architecture based on WSRF wh ere OGSA works directly over WS.
5. Semantic Web Semantic W eb i s defined as an ext ension of Web technology in which the information has its own meaning. It makes possible for the web to understand and satisfy the requests o f p eople a nd m achines using web c ontent. Besides, Seman tic Web allows ob taining effective discovery, i ntegration, a utomation an d re use of applications [10]. Semantic W eb was orig inally p roposed b y Ti m Bern ersLee, Jam es Hen dler and Ora Lassila in 20 01, with t he objective of ex tend to t he Web so me o f th e in telligent behavior of humans [7]. Semantic W eb technologies are bas ed on open standards. If a computer understands the semantics of a document, it does not only interpret t he series of ch aracters th at m ake up that document but also understands the meaning of the information in the document. The technologies that integrate the Semantic Web are: Uniform Res ource Identifier (URI): It is a global naming scheme. Resource De scription Fr amework (RDF): It is a standard syntax for describing data. RDF is a specification that defines a model for representing the world, and syntax for serializing and exchanging that model [15]. RDF Schem a: It i s a dat a t yping m odel f or R DF which allows creating properties and classes. It coul d also be defined as a vocabulary f or desc ribing properties an d classes of resources based on R DF a nd ag gregating semantics descriptions [16]. Web O ntology La nguage (O WL): It is designed to be used by ap plications t hat need t o process t he c ontent of information rath er t han ju st p resenting i nformation to humans. It a dds m ore v ocabulary t o desc ribe p roperties and classes than RDF or RDF Schema [15, 16].
Fig. 1 (a) Architecture based on OGSI. (b) Architecture based on WSRF
OWL-based Web Service (OWL-S): It supplies W eb service providers with a set o f m arkup lang uage constructed t o d escribe th e properties an d cap abilities of their WS in a computer-interpretable form [17].
Figure 1 (a) shows the approach based on OGSI where WS are m anipulated t o s atisfy t he nee ds of G rid C omputing aggregating functions and new specifications and creating what we have called GS, wh ile part (b) shows the
Semantic t echnologies re present meaning t hrough the use of ontologies[18]. Ontologies, in terms of Semantic W eb, are defined as meta-data schemas that provide a controlled vocabulary of concepts, each of them with a complete and
(a)
(b)
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org structured seman tic [ 19, 20 ]. Th ey ar e descriptions of concepts f or the p urpose of allowing kn owledge sharing and pr ovide m achine reada ble and un derstandable dat a [16]. OWL W eb O ntology Language w orks wi th ontologies defining and instantiating t hem. OW L is also used t o represent th e meaning of term s in vo cabularies and the relationship existing among these terms [16]. The use of Se mantic W eb in o ur applications has its benefits. The Semantic Web provides the ability to tag all content on the W eb, descri bes each piece of information, and gives semantic meaning to each resource. Thus, search engines become more e ffective, a nd users ca n fi nd t he precise i nformation t hey are l ooking f or. Organizations providing services can tag se rvices with m eaning; usi ng software a gents base d on t he Web, t hese s ervices ca n be found and used dynamically [21].
4
necessary to invoke a service, i nformation about companies that own the service, and business entities [22]. If w e ad d seman tics to UDDI , we cou ld also obtain good results in our search. The Sem antic Web ap proach t o GS, no t only facilitates services discovery, but could also improves th e selection , invocation, execu tion, co mposition and in teroperation of Grid/Web Serv ices as it does fo r serv ices d iscovery[23] through the application of Semantic Web technologies. Consequently, we ca n observe t hat t he use o f Semantic Web and ontologies provide an appropriate mechanism to represent GS an d to enhance their discovery, composition, selection, invocation and interoperation.
6. Conclusions
As mentioned above, through WSRF, GS can be seen l ike WS. Moreover, we ha ve to rem ember that WS are described using WSDL definitions and advertised in UDDI registries.
In t his pa per, we have defined Grid Ser vices as so ftware components whic h pe rmit us to access a set of Gri d resources s uch as data s ources, hi gh perform ance equipment and computational resources. Grid Services a re an a pproach to Web Se rvices; t hey have t heir o wn architecture, OGSA, a nd i nfrastructure, WSRF, wh ich allows us to manipulate Grid services as Web services.
The capability of a WS is defined through several aspects, such as t heir a utomated disc overy. The c urrent disc overy mechanism sup ported by UD DI i s not ade quate for automated d iscovery. Its m ain li mitations are th e lack of semantics in the discovery process and that UDDI does not use in formation i n the s ervice descriptions d uring discovery; this reduces the effectiveness of UDDI [18, 22]. Because GS are treated as WS through the WSRF standard, these limitations also affect them.
Furthermore, we have demonstrated that Grids have taken advantage of the technologies and standards developed for Web Services such as WSDL, UDDI and SOAP, and it has begun t o use Semantic W eb - defined as a n e xtension of the Web tech nology where in formation h as its own meaning- and its techno logies (o ntologies, R DF, OWL, RDF-Schema, OWL-S) to aggregate semantics to the Grid Services descriptions a nd enha nce t heir discovery, composition, selection, invocation and interoperation.
The semantic discovery of Web/Grid services is achie ved adding sem antics t o WSDL and U DDI, a nd t hen we ca n find th e requ ired serv ices u sing sem antic matching algorithms.
These are si gnificant as pects because a Grid Com puting environment needs t o have an efficient search service a nd quick res ponse t ime i n addi tion t o ade quate and acc urate results. Su bsequently, thro ugh th e conv ergence between Grid and Semantic Web these needs can be fulfilled.
6. Discussion: Semantic Grid Services
In order to add semantics to WSDL of a s ervice and find relevant operations, t hese o perations m ust be m apped t o concepts i n a ppropriate O WL ontologies. Th e use o f ontologies allow to represent service descriptions (WSDL) in a machine-interpretable form like OWL-S [17, 18]. Additionally, WSDL c ontains m essage p arts de fined a s input and output op eration parameters, th us, we can add ontologies and RDF con cepts to an notate message parts in WSDL, and later we can perform more effective searc hes for Grid/Web Services and obtain better results. On t he ot her ha nd, we ca n i mprove semantic servi ce discovery and storing sem antics in t he ex isting stru ctures of UDDI. UDDI contains the technical information that is
References
[1] I. Foster, C. Kesselman, and S. Tuecke, "The Anatomy of the Grid: Enabling Scalable Virtual Organizations”, Proceedings of the Intern ational Journal of S upercomputers Applications, Vol. 15, 2001. [2] I. F oster, J . N ick, C. K esselman, and S . T uecke, "Th e Physiology of the Grid: An Op en Grid Serv ices Architecture for Distributed Systems Integration", 2002. [3] I. Foster, and C. Kesselm an, "Globus: A Metacomputin g Infrastructure Toolkit" , Int. Journal of Su percomputer Applications, 1997, pp. 115-128. [4] E. Al-Masri, and Q. Mahmoud, "Discoverin g the Best Web Service: A Neur al Network-b ased S olution", P roceedings of the 2009 I EEE International Co nference on S ystems, Man, and Cybernetics, 2009, pp. 4250- 4255.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org [5] M. Muhammad, L. Yu an, and Z. Jianqiu , "Web 3.0: A real Personal Web", 2009 Third International Conference on Next Generation Mobile Applications, Services and Technologies, 2009, pp. 125 – 128. [6] O. Lassila, and J. Hendler , "Embracing W eb 3 .0”, I EEE Internet Computing, Vol. 11, 2007, pp. 90-93. [7] S. Chun, J. Geller, and Y. Jung, "Toward the Semantic Deep Web", Computer, Vol. 41, pp. 95-97, 2008. [8] G. Alonso, F. Casati, H . Kuno, and V. M achiraju, "Web Services: Con cepts, Archit ectures, and A pplications", Springer Verlag, 2004. [9] N. Park, and G. Lee, "Agent-based Web Services Middleware", 2003. [10] C. Goble, D. DeRour e, "The Grid: An Ap plication of th e Semantic Web", SIGMOD Record, Vol. 31, pp. 65-70, 2002. [11] M. Younas, K-M. Chao , R. Anane, S. Y. Yan, P. J. Lovett, and A. N. Godwin, "Grid S ervices M ediation", P roceedings of the 19th Intern ational Conferen ce on Advanced Information Networking and Applications (AINA’05), Vol. 1, pp. 879-884, 2005. [12] I. Foster, K. Czajkowski, D. Ferguson, J. Frey, S. Graham, T. Maguire, D. Snelling , and S. T uecke "Mo deling an d Managing State in Distributed Systems: The Role of OGSI and WSRF", Proceedings of the IEEE, Vol. 93, pp. 604-612, 2005. [13] D.Tal ia, " The Open Grid S ervice Archit ecture: W here th e Grid Me ets t he We b", IE EE Internet Computing, Vol. 6 , 2002, pp.67-71. [14] W. Liu, Z. Liu, and W. Tong, "Agent-Oriented Modeling for Grid Service", Proceedings of th e Sixth International Conference on Machine Learning and Cybernetics, 2007. [15] V. Muppavarapu, and S. Chung, "Semantic-based Access Control for Grid Data Resources in Open Grid Services Architecture - Data Ac cess an d Integration (OGSA- DAI)", 2008 20th IEEE International Confer ence on Tools with Artificial Intelligence, Vol. 2, 2008, pp. 315- 322. [16] S. Sane,an d A. Shirke, "Generating OWL Ontologies from Relational Da tabases for the Sem antic W eb", International Conference on Advances in C omputing, Communications and Control, 2009, pp. 157-162. [17] D. Mar tin, M. Burste in, D. Mcd ermott, S. Mc ilraith, M.Paolucci, K . Sycara, D. L. Mcguinness, E. Sirin, and N. Srinivasan, "Bringing Semantics to Web Services with OWLS", World Wide Web, Vol. 10, 2007, pp. 243-277. [18] K. Sivashanmugam, K. Verma, A. S heth, and J . M iller, "Adding Semantics to Web Services Standards", Proceedings of The 2003 I nternational Co nference on Web Services (ICWS'03), 2003, pp. 395-401. [19] A. M aedche, and S . S taab, "Onto logy Learning for the Semantic Web", IEEE Intelligent Systems, Vol. 16, 2001, pp. 272-279. [20] J. Hendler, "Agents an d t he Se mantic We b", IE EE Intelligent Systems, Vol. 16, 2001, pp. 30-37. [21] M. d 'Aquin, E. Motta, M. Sabou, S. Angeletou, L. Gridinoc,V.Lopez , and D.Guidi, "Toward a New Generation of Sem antic W eb Applic ations", IEE E Int elligent S ystems, Vol. 23, 2008, pp. 20-28. [22] S. Banerjee, S. Basu, S. Garg, S. Garg, S. Lee, P. Mulla n, and P . S harma, "Scalable Grid Service Discover y Based on UDDI", ACM I nternational Conf erence Proceeding Series, Vol. 117, 2005, pp. 1-6.
5
[23] S. McIlraith, T. Son, and H. Zeng, "Semantic Web Services", IEEE Intelligent Systems, Vol. 16, 2001, pp. 46-53. Belén Bonilla-Morales is a student of the Master of Science in Information and Communication Technology at the Technological University of Panama. She was awarded a Bachelor’s degree in Engineering and Computer Science from the Technological University of Panama in 2009. Her research interests include Software Engineering, Semantic Web, Grid Computing and other topics. Xavier Medianero-Pasco is a student of the Master of Science in Information and Communication Technology at the Technological University of Panama. He was awarded a Bachelor’s degree in Engineering and Computer Science from the Technological University of Panama in 2009. His research interests include Aspect Oriented Programming Framework, Grid Computing and other topics. Miguel Vargas-Lombardo is a professor at the Technological University of Panama, PhD awarded by the Polytechnic University of Madrid.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No. 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
An Enterprise Architecture Framework for Mobile Commerce Khawar Hameed1, Hanifa Shah2, Kamran Ahsan3 and Weijun Yang4 1234
Faculty of Computing, Engineering and Technology, Staffordshire University, Stafford, United Kingdom
Abstract
The increasing deplo yment o f m obile t echnologies across industry sectors is creating fertile ground for organisations to exploit n ew rev enue streams gene rated from a pplications th at exploit the mobile ecos ystem. M-Commerce has been recognised as a key driv ing for ce of n ext g eneration computing and industry analysts such as IDC have predicted revenue growth arising from m-commerce to far exceed US$27 billion by the end of the d ecade [1]. Mobilit y has, without doubt , underpinned the current wav e and gen eration of computing s ystems resulting in the concept an d prac ticality of m obile s olutions becom ing embedded as n atural or inh erent ones that su pport th e d aily functions of individuals and corporations. This paper explores an approach to encapsulating the m-commerce ecosystem through the perspective of an enterprise architecture framework.
Keywords: Mobility, Mobile Commerce, Mobile Applications, Mobile Innovation, Mobile Services
1.Introduction The phenomenon of mobility has driven an ev olution and revolution of t echnologies and ne w way s of w orking, exploiting the practice of anytime, anyplace, any how computing whilst g aining critical mass as a research discipline and paradigm - the study of which has provided a gro unding for co nceptual and th eoretical p erspectives that allo w framin g an d discussion of m obility [2 ] [3 ]. Mobile c ommerce has created a wide ra nge of business opportunities - the sp ectrum o f wh ich in cludes t he transformation o f mobile n etwork op erators to go beyond the prov ision of co re voice serv ices t o th e d elivery of value added services and service a ggregation that provide comprehensive co nsumer pac kages [4], t he emergence of business m odels t o develop an d deploy m obile applications (such as the Apple ‘App’ Store, the O 2 Litmus programme, and the Sony Eri csson Deve loper Worl d – most of w hich are based on sha red re venue distribution through sales th rough th e respective ho sting ch annels. Furthermore, ope n pr ogrammes and pl atforms such as Android ( Google) are l ikely t o c ontribute t o t he enlargement of the developer comm unity that seeds the growth of mobile applications. W ith the sup ply chain o f mobile computing and commerce, a range of opportunities
exist for con tributing p arties to co llaboration t o prov ide value-added s olutions based on new rel ationship t ypes, rules and ecosystems [5] further enhancing the composite capability and reach of service providers into new market areas. Bey ond se rvice providers, t he ea se of access , penetration, a nd diffusion of m obile t echnologies has enabled i ndividual sect ors a nd organisations t herein t o apply th e co ncepts and practice o f m obility to create innovative domain-specific m-commerce a pplications t hat leverage the specific nua nces ass ociated with t hose domains resulting i n value-added s olutions f or e nd-users and n ew revenue-generating op portunities fo r business. Examples of these i nclude mobile location-based tourism, travel and na vigation sy stems, m-ticketing and b ooking applications. Beyond th e co re comm ercial im perative of increased revenue generation, th e app lication of m obile computing an d principles o f m -commerce also underpi n the de velopment of pri vate and pu blic sect or mobile applications t hat aim t o red uce ope rating cost s, e nhance efficiency an d p rovide bet ter platforms f or e ngaging t he end-user population. Baso le [ 2] pr esents an En terprise Mobility Co ntinuum th at fra mes mobile solutions fro m point-specific solutions t o t hose t hat diffuse acro ss e ntire organisations to creat e st akeholder value. Acr oss a nd throughout the co mplex m-commerce ecosystem the question arises reg arding the construction of m -commerce solutions and how these are best approached. The i ncreasing at tention t o b usiness u nderpinned by mobility, mobile serv ices, mobile applicatio ns, and technologies has become a major driver for the development of m-commerce systems. This growing trend has bec ome a foc us for a signi ficant num ber of organisations. Thi s pape r pr oposes that in devel oping mcommerce sy stems orga nisations need t o est ablish a n enterprise architecture for m -commerce. The rationale for this is roo ted in th e n eed to d evelop a h olistic and integrated view of st rategic di rection relating t o m commerce which will enable a coordinated and controlled approach that re duces com plexity an d y ields effectiv e systems based on the structur ed integration of ser vices, practices a nd technology re sources. I n doi ng so, t he potentially complex universe of discourse associated with
6
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No. 5, May 2010 www.IJCSI.org mobility i s har nessed t o produce a nd organisational asset to drive the development of m-commerce. The ne xt section proposes a f ramework t o est ablish s uch an enterprise architecture fo r m obile co mmerce. Firstly , an enterprise architecture framework for mobile commerce is presented followed by a brie f view of associated issues concerned with method.
7
scope the need for a c omprehensive and consistent approach i n m odelling these system s bec omes of paramount im portance.” B asole [2 ] reco gnises th at businesses are “just beginning to recognise the importance and po tentially tran sformative i mpact o f en terprise mobility.” Gi ven t his our ap proach t o ap plying EA principles t o m-commerce appea rs well-grounde d. A proposed E nterprise A rchitecture F ramework for m obile commerce is shown in Figure 1.
2. An Enterprise Architecture (EA) Framework for Mobile Commerce An EA fram ework provides th e b asis or t emplate for the creation and estab lishment o f ente rprise architecture . Zachman [6 ] is cred ited with d eveloping the d iscipline of enterprise ar chitectures a s a concern for both r esearchers and practitioners. An enterprise architecture framework is essentially a meta-construct used to define the scope of the associated architecture and how the areas of t he architecture re late to each other [7]. An a rchitecture ca n be c onsidered anal ogous t o a blueprint or pl an of a building structure, w here di fferent pe rspectives m ay exist and eac h perspective c ontains st ructures that dem onstrate inter-relationships based upon some pr edefined co nstraint and yield a s olid foundation and approach upon which the building is constructed. Generally s peaking, t he E A framework defines t he sc ope of th e resulting arch itecture, wh ich typ ically in cludes a business view, in formation in tegration, app lication-level views, and technology infrastructures. Definitions of each view m ay include m ore refi ned c onstructs an d relationships at a lower level of granularity. The application of an EA approach is considered relevant and a ppropriate since the e cosystem with in wh ich t he development of m-commerce solutions occurs comprises a set o f in terrelated p erspectives b ased upon the in tegration of m obile de vices, t echnologies, a nd business processes [8]. T herefore, an E A framework f or m obile com merce can be c onsidered t o a ddress, at least, the sc ope of architecture cove ring the busi ness l evel/view, t he application level and the technology infrastructure level. In each level of our propos al, core c omponents have been identified. The se are: a bu siness m odel of m-commerce (business l evel), s upply cha in of m -commerce (s upply chain lev el), m-commerce ap plications (application lev el) and tec hnology infra structure for m-commerce (technology l evel). Our research ha s s hown t hat integration of core EA ap proaches with m -commerce is relatively sp arse, wh ereas literatu re in both co ntributing areas is sub stantial. Furthermore, Leist & Zelln er [9] state that “as informatio n systems grown in co mplexity an d
Fig. 1 An Enterprise Architecture Framework for Mobile Commerce
2.1 First Level: M-Commerce Business Models Figure 1 s hows the first (t opmost) level in our Enterprise Architecture fram ework for m -commerce relating t o business models. This level is identified as the first level in the EA framework and provides a desc ription of the roles and relationships of an organisation, its customer, partners and sup pliers and stak eholders, as w ell as th e flows of goods, i nformation a nd m oney bet ween t hese pa rties a nd the main benefits for those involved [10]. The stakeholder transactional m odels are b ased upon tho se pr esented by Coursaris & Hassanein [11] and ar e d ivided in to fou r models (i ) wi reless B usiness-to-Consumer ( WB2C) m odel; (ii) a wireless Consumer-to-Business (WC2B) model; (iii) a wireless Consumer-to-Consumer (W C2C) model; and (iv) a wireless C onsumer-to-Self (WC2) m odel. These m odels mainly d escribe bu siness activ ities between th e contributing parties, and the specific nature of t hese. T he abstraction of these m odels provides a ba sis fo r higherlevel u nderstanding of t he s pectrum of st akeholders an d relationships at the topmost level of the architecture. Other business m odels of mobile co mmerce attem pt to address the complexities associated with the m -commerce ecosystem com prising di fferent participants (suc h a s mobile consu mers, net work ope rators, se rvice p roviders,
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No. 5, May 2010 www.IJCSI.org application developers, c ontent p roviders and t echnology providers) a nd enc ompass th e types of services and sources of profits. In addition, models describe operations and pro cesses relatin g to mobile g rowth, value-added benefits, revenue models and return on investment, and the transfer of be nefits ac ross stakeholders for m utual gai n [12]. During the de velopment of ou r fram ework we re viewed models broa dly associ ated with t he business m odels fo r mobile in formation, m obile ad vertising and for m obile office work. Typical c haracteristics of these were associated with r elationships bet ween c ontent p roviders, network ope rators and service providers f or di rect gai n through revenue opportunities from mobile consumers and shared distribution amongst stakeholders and participants. Interestingly, o ne specific find ing was th at, “mobile consumers re present the only sustaina ble revenue s ource for participants” [13]. From the characteristics identified, it is seen that business models fo r mo bile co mmerce are helpful in en abling organisations t o in crease th e p ossibilities fo r in creasing revenue and profit, and enh ancing competitiveness. Taking Vo dafone G roup Plc [14] as a typical exam ple rather co mplex bu siness activities can be observed wh ere the b usiness acts as wireless network o perator prov iding information tran smission services for its cu stomers wh ilst also acting as a content or service providers of specific service conte nt for its customers s uch as m usic downloading, online gaming, e-m ail an d lo cation-based information. In th is and other sim ilar cases th e business models ap pear no t to be m utually ex clusive (e.g . si ngle and independent WB2C, WC2B, WC2C, WC2 business models) [11]. Th us, e ncapsulating t he diverse st akeholder groups and ass ociated m obile inform ation needs can be seen as providing co mpetitive ad vantage and m arket diversification.
2.2 Second Level: The Supply Chai n of Mo bile Commerce Figure 1 shows t he second level in t he En terprise Architecture fram ework for mobile commerce and is concerned with th e su pply ch ain. The m aterialisation o f business m odels fo r m -commerce dep ends o n a c omplex chain of business relationships between participants of the supply ch ain [12]. The supply ch ain for m-commerce is therefore id entified as th e seco nd an d supporting lev el i n the enterprise architecture framework. This supply c hain for m-commerce can be seen as rooted in m obile tele com markets, with in wh ich a v ariety of participants p ossess resources, perform acti vities, and are in rel ationships t hat are established or ev olving i n
8
delivering an end-to-end service [15]. However, given the increasing diversity o f ap plications, services an d associated information delivery, this supply chain is b eing extended t o include a s pectrum of market areas which, when ag gregated provide value-added se rvices t o t he mobile user. The c ore ele ments of t he m -commerce supply c hain include m obile co mmerce p articipants, m obile co mmerce resources and m obile co mmerce activ ities. Mo bile commerce participants a re a major dri ving force be hind the m -commerce s upply chain [13]. Mobile c ommerce participants fu ndamentally in clude m obile co nsumers, wireless n etwork prov iders, co ntent provide, serv ice providers, a pplication d evelopers and technology providers – al l of whom engage i n a business-oriented relationship and, in t he specific case of m -commerce, one that is fo cussed on co mmercial gain for s ervice delivery participants. For ex ample, th e shared /distributed revenue approach entails the wireless network providers delivering part of t hat r evenue t o ot her pa rticipants i n t he s upply chain – suc h as c ontent providers a nd ot her service providers.
2.3 Third Level: Mobile Commerce Applications The third level of the framework is concerned with mobile commerce applications. Becau se a n organisation use s mobile ap plications to su pport and d eliver its bu siness models th rough th e supp ly ch ain, m obile ap plications ar e seen as the tangible end-user vehicles that mechanise and enable t he m -commerce trans action. M obile app lications are therefore identified at the third lev el in th e En terprise Architecture framework for m-commerce. M-commerce applications can be broadly categorised a s communication ap plications, inform ation ap plications, entertainment applications and c ommerce a pplications [11]. Five m ain app lication typ es were i dentified: m obile ticketing, m obile ad vertising, m obile in formation, m obile banking and mobile office applications. These were found to supp ort the k ey business m odels and im peratives (revenue g eneration or co st r eduction) an d w ere constructed using th e supp ly ch ain t o d eliver an aggregated service. T he cons tantly d eveloping lan dscape of m obile t echnologies, an d more specifica lly application capability, rai ses a pro position of re-aligning business models and supply chains to fully leverage the potential of that ch ange - thus sug gesting a co mmercially syn ergistic relationship (i n th is case, the app lications and techno logy forcing a re -evaluation of the business models and s upply chain) .
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No. 5, May 2010 www.IJCSI.org
2.4 Fo urth Level: The Technolog y Infrastructure for Mobile Commerce The f ourth l evel o f t he f ramework i s c oncerned with technology infrastructure for m-commerce. This includes wireless c ommunication t echnology, wireless m iddleware technology, i nformation e xchange t echnology, wireless network & a pplication protocols a nd m obile se curity technology. Essen tially, these are the core technol ogical components and infrastructures that enable mobile users in their environments. These t echnologies su pport upper levels of the f ramework (m -commerce business models, m-commerce supply chain, a nd m -commerce applications).
that would enable scope for interpretation and discussion. Three org anisations (anonymity maintained) p articipated in th e surv ey. Qu estions were ask ed in t wo p arts (EA Framework for M-C ommerce & E A M ethod for MCommerce) and s ought vie ws on th e fo llowing areas and propositions:
3.1 The Business Model for M-Commerce A1: The p ositioning o f t he busi ness m odel for mcommerce at the first (topm ost) level in the EA Framework A2: The role of a business model for m -commerce to describe how organisations create and real ise genuine benefit from mobile commerce activities.
3. Method In order to materialise a s pecific arc hitecture from the enterprise architecture framework an associated method is required. The method provides a step-by-step description of how to estab lish the architecture [7]. In context of the proposed framework, a five-step a pproach i s proposed based u pon five ge neric st eps [ 16] ada pted t o s uit t he framework-specific nee ds. Our c urrent re search f ocuses on the development of the m -commerce e nterprise architecture framework and th e method associated with this i s i n em bryonic f orm. Neve rtheless, vi ews an d expectations are th at this will focus on in ternal knowledge audits t o determine or ganisational rea diness and m obile strategies, a series of data collection m ethods that aim to elicit th e typ e o f ex isting of new app lications th at can be mobilised in co ntext of m -commerce, th e sk ills b ase, an d the technology/resource levels within the organisation, and levels of innovation. Other aspects are expected to address appropriate business planning tools, modelling techniques and notations t hat appl y across t he f our l evels of t he framework and in corporate the m odelling o f m obility i n context of mo bile co mmerce opp ortunities. Techn ical limitations of m obile d evices and wireless communications, business c oncerns a nd l egal c onstraints complicate the practical use of m obile co mmerce [17] . Therefore, t hese al so p rovide t he im petus of developing appropriate methods that addr ess t hese concerns. In al l these cases o rganisations m ay have e xisting a pproaches and techniques that might be used a part of the method.
A3: The division of an m-commerce business model into a W ireless B usiness-to-Consumer model, Wireless Consum er-to-Business m odel, Wireless Consumer-to-Consumer m odel, and Wireless Consumer-to-Self model [11]. A4: The c omposition of a business m odel for m commerce to include th e so urce o f profits, mobile services and supply chain participants. A5: The choice of novel or more recent business model approac hes over t raditional busi ness models for m-commerce.
3.2 The Supply Chain for M-Commerce A6: The positioning of the supply chain model for mcommerce at the second level in the EA Framework. A7: That a good supply chain for m -commerce can streamline business processes for m-commerce. A8: That the supply chain for m -commerce includes m-commerce participants, m-commerce resources and m-commerce activities. A9: That the m -commerce participants are a major driving fo rce b ehind th e su pply ch ain of m commerce.
3.3 M-Commerce Applications A10: The positioning of m -commerce applications at the third level in the EA Framework.
4. Validation To validate t he pr oposed fr amework a qu estionnaire was developed t o t est t he o verall co nstruct o f t he f ramework, its d irect relevan ce to businesses, and to seek the views that m ight be incorporated i n i ts re finement. T he s urvey aimed to elicit initial and relatively informal feedback and as su ch was co nstructed to provide qu alitative feedb ack
9
A11: M-commerce applications are influenced by the business m odel and supply chain for m-commerce.
A12: M-commerce applications need be supported by reliable m-commerce technology
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No. 5, May 2010 www.IJCSI.org
3.4 Technology Infrastructure for M-Commerce A13: The p ositioning of t echnology i nfrastructure fo r m-commerce at th e fin al l evel (b ottom) in th e EA Framework. A14: T he rol e of m obile comm unication technol ogies in offering message or wireless data capability. A15: That wireless middleware technology supports the development and o peration of m-commerce applications. A16: Information exchange technologies are necessary for m-commerce applications and mobile devices. A17: That wi reless net work and ap plication pr otocols can support the delivery of web-based information.
3.5 EA Method for M-Commerce B1: The p ositioning o f data col lection ( knowledge audit) as a first step in an EA method for m-commerce.
Table 1: The total feedback score - Organisation 1
B2: Data co llection sh ould comprise identification of m-commerce components and also possible obstacles or restrictions in m-commerce. B3: That defi ning t he p urposes of an E commerce is defined as the second step.
A for m -
B4: That esta blishing an E A (sc hema) for m obile commerce is defined as the third step. B5: T hat an E A fram ework of m -commerce needs be modelled by software tools. B6: That tool selection is defined as the finall step.
4. Results A 3 scale s coring system was us ed a nd res ponses constrained to Accept (3 ), Neither/Neutral (2 ), a nd Not Accepted (1 ). All t hree participants responded and t he summarised resu lts for each p articipant are tabu lated as shown.
Table 2: The total feedback score - Organisation 2
10
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No. 5, May 2010 www.IJCSI.org
11
financial, organisational, t echnological, a nd s ocial boundaries. It is th is th at p rovides t he i mpetus and rationale t o f ocus on adopting an e nterprise arc hitecture framework ap proach t o govern t he development o f m commerce syste ms and with a vie w to reducing complexity. It is also th is th at ai ms to en able organisations to adapt rapidly to mobile technologies, reap the affordanc es create d t hrough techn ology adop tion, remain co mpetitive – yet throug h a structured and framework-oriented ap proach th at prov ides a b aseline for pro-active t ransformation r ather t han a re-active a nd potentially chaotic and fragmented approach.
Table 3: The total feedback score - Organisation 3
All 3 res pondents sco red hi ghly re garding t he b usiness model com ponent of t he p roposed f ramework, i ndicating its ap propriate a nd relevant positioning. Si milarly, responses regarding the supply chain model component of the framework sco red highly. Th is ag ain was in terpreted as a po sitive response reg arding its ov erall co ntext and positioning. Th e scoring for m-commerce applications as part o f the fra mework was slightly lo wer, although t his was still d eemed a p ositive respo nse. Fin ally, th e technology infrastru cture l evel sco red highly, ag ain indicating an ov erall positive respon se. Th e overall interpretation was t hat th e fra mework proposition was an appropriate an d useful vehicle f or t he development o f mcommerce within organisations. With respect to the method, all responded again scored this aspect highly thus indicating its potential relevance, value and use.
5. Conclusions Mobile technology is continuing to play a si gnificant role in p roviding efficient and ef fective me ans for organisations to br oaden t heir re venue st reams, and enhancing competitive positioning. There is no doubt that the m obile ec osystem creat es many p ossibilities fo r the development of innovative solutions that create real value for en d users whilst meeting real fi nancial i mperatives of service pr oviders and s takeholders in th e m- commerce supply c hain. However, the m obile ecosystem is potentially co mplex an d th e un iverse of d iscourse created contains m any i nterrelated c omponents t hat span ac ross
This paper has propo sed an i nitial en terprise arch itecture framework for m obile com merce that aim s to provide practitioners and researchers a platform for considering the development of m-commerce systems, and one which aims to influence both philosophical and practical approaches to building m -commerce system s. In itial resp onse, al beit being b ased on li mited d emographics, has b een positive and t his p rovides m otivation for future wo rk. T he framework proposition co mprises four lev els, each of which dr aws up on specific nua nces ass ociated with the mobile envi ronment and eac h of which i s constructed i n context of other layers – thus demonstrating a frameworkoriented approach. Acknowledgments We wou ld lik e to th anks th e th ree o rganizations (anonymity maintained) fo r th eir p articipation i n th e survey.
References [1] Mahatan ankoon, P ., Wen H.J. , & Lim,B. (2 005) ConsumerBased m-commerce: Exp loring Consumer Perception of Mobile Applications. Computer Standards & Interfaces, 27(4), 347-357. [2] Basole, R ahul C ., Enterprise Mobi lity: Researching a New Paradigm. Infor mation Knowle dge S ystems Management 7 (2008) 1-7. [3] Kakihara, M ., Sorensen, C. (2001) Expanding the ‘Mobility ’ Concept. ACM SIGGROUP Bulletin. 22 (3) 33-37. [4] Sey bold, Andrew M., The Conv ergence of Wireless, Mobility, and th e Intern et and i ts Relevan ce to Enterprises. Information Kn owledge S ystems Management 7 (2008) 1123. [5] McDowell, M., Business M obility: A Chan ging Ecosystem. Information Kn owledge S ystems Management 7 (2008) 2537. [6] Zachm an, J.A., A F ramework for Infor mation S ystems Architecture. IBM Systems Journal, Vol 26 (3) 1987. [7] Scott A. Bernard, An Introduction to Enterprise Architecture, Second Edition, 2006.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No. 5, May 2010 www.IJCSI.org [8] N.Band yo-padhyay, E-Com merce: Contex t, Concep ts and Consequences, Published b y McGraw-Hil l Educa tion, Second Edition, 2002. [9] Zellner, S. L. A. G. (2006 ). Evaluation of current architecture frameworks. A CM, SAC' 06, April, 23-27, 2 006, Dijon, France. [10] Giovanni Camponovo and Yv es Pigneur, Business Model Analysis Applied To Mobile Busine ss, The University of Lausanne, 2004. [11] Coursaris & Hassanein, Understanding m-commerce: A Consumer-centric Model, Degroote School of Business, McMaster University, 2006. [12] WAN G Yan & GAO Yufei, Re search on the Value Chain and Busin ess Models for Mobile Commerce, Fudan University, 2004. [13] Brian E.M ennecke and Tr oy J.Str ader, M obile Commerce: Technology, Theor y, and Applications, Lowa State University, Published by Idea GroupInc, 2003. [14] Vodafone, About Vod afone Power to You , from http://www.vodafone.com/start/about_vodafone, accessed November 13, 2009. [15] Nan Si Shi, Wireless Communications and Mobile Commerce, Published in the United Kingdom by Idea Group Inc, 2004. [16] Rob.C.Tho mas, A Practica l Guide for Developing an Enterprise Arch itecture, Federal Chief Information Officer Council, 2004. [17] Keng Siau, Ee-Peng Lim, Zixing Shen, Mobile Commerce: Current States and Future Trends, University of NebraskaLincoln, USA, 2006. Khawar HAMEED is a Principal Lecturer in the Faculty of Computing, Engineering & Technology at Staffordshire University. His research is in the area of mobile and flexible working, enterprise architectures to support mobile information systems, and mobile learning. He has been a key driver in the adoption of mobile computing and technology within the Faculty's portfolio and has helped drive the development of undergraduate and postgraduate degrees in this technology area. He has contributed extensively to the development and delivery of externally funded projects and academic-industrial collaborations in mobile/wireless technology that aim to develop and enhance the collective intellectual capital that supports the growth of mobile and wireless systems as a discipline both within academia and in industry. Hanifa SHAH is Professor of Information Systems and the Director of the Centre for Information, Intelligence and Security Systems at Staffordshire University. Her research interests and publications are in the areas of Knowledge Management, Information Systems Development, Enterprise Architecture, IT professionalism, and Mobile Technologies in Patient Processes. Kamran AHSAN has an MSc in Mobile Computer Systems from Staffordshire University and MCS (Masters in Computer Science) from the University of Karachi. Kamran is a PhD researcher and lecturer in FCET (Faculty of Computing, Engineering and Technology) and a web researcher in the Centre for Ageing and Mental Health, Staffordshire University, UK. He has published several papers and has been involved in a number of UK funded research projects including KTP, NHS Trust, and Innovation Vouchers. He is Visiting Faculty at the University of Karachi. He is a consultant to businesses in IT applications, software development and web tools. His research interests are in mobile technology applications in healthcare including knowledge management.
12
Weijun YANG is a postgraduate candidate in Computer Science in the Faculty of Computing, Engineering & Technology at Staffordshire University. His focus of work is in the area of enterprise architectures to support mobile commerce systems. Previously he worked as a technical consultant in performance and functional testing of IT systems, successfully developing and implementing performance optimization programs. His research interests are in the development of enterprise architectures and performance metrics for evaluating the function and effectiveness of mobile commerce solutions.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
13
Performance Comparison between Active and Passive Queue Management Arash Dana1 and Ahmad Malekloo2 1
Department of Electrical Engineering, Islamic Azad University, Central Tehran Branch Tehran, Iran 2
Department of Computer Engineering, Islamic Azad University, Arak Branch Arak, Iran
Abstract The queue m anagement al gorithm, whi ch is applied to a router, plays an im portant role in providing Quality of Service
(QoS). In this paper we pres ent a simulation-based performance evaluation and compare two popular queue management methods; Random Early Detecti on (RED) and Drop Tail, in terms of queue size, queuing de lay, queuing delay variation, packet drop rate and bandwidth utilization. W e also study the effect of buffer behavior on each one of these QoS measurements, as well as considering the effects of using ECN on RED Performance. Simulation results indicate that RED perform ed better related to Drop Tail in terms of queuing delay, queuing delay variation, and (lower) pack et drop rate. Furthermore, using ECN improves RED's bandwidth usage and packet drop rate. Also by measuring the router CPU utilization, we indicate that the difference between active a nd passive queue management algorithms in terms of CPU utilization, is m ostly depend on the buffer size and the network traffic load. Keywords: Queue Management, Random Early Detection, QoS, Congestion Control.
1. Introduction Recently, the exponential growth of the Internet and the use of new services such as e-business, voice over IP (VoIP) an d m ultimedia ap plications h as b een risen th e need of supporting Quality of Service (QoS) requirements. The queue m anagement al gorithm, whi ch is applied to a router, plays an important role in all of QoS measurements (include delay, jitter, bandwidth and packet loss). Queuing management algorithm is responsible for accepting the arriving packets or not accepting them and consequently it directly affects th e p acket lo ss q uantity parameter. The queue management algorithm also should ideally keep t he queue occupat ion l evel as l ow as possible, to ensure low delay. Ho wever, to en sure m aximum u tilization o f th e outgoing l ink, t he queue shoul d never be empty. Furthermore, m aintaining q ueue stab ility is important as some applications are sensitive to jitter.
As a resu lt, u ntil n ow m any q ueue m anagement schemes have been introduced in th e literatu re in o rder to av oid congestion and improve the QoS in the network. But there are only a few al gorithms whi ch st andardized and implemented. Drop Tail i s t he m ost wi dely used queue manage algorithm in today's IP networks due to its simple implementation [1]. While Random Early Detection (RED) [2], an act ive queue m anagement schem e, i s avai lable in most routers nowadays, it is often disabled by default [3]. RED has been recom mended by t he Internet Engineering Task Force (IETF) as the default active queue management scheme for t he next generation networks [4, 5]. Almost all other active queue management schemes are presented as improvements on RED. The main aims of this paper are i) to compare two popular queue management methods, RED and Drop Tail, in terms of queue size, queuing del ay, queui ng del ay vari ations, packet drop rate and bandwidth utilization; ii) co nsidering the effects of using ECN m echanism on RED algorithm Performance; iii) considering the effect o f buffer behavior on QoS m easurements; iv) to com pare and find the difference of CPU u tilization b etween activ e an d p assive queue management algorithms. The Rest of the paper is organized as follows. In section 2 we will review Drop Tail an d RED alg orithms briefly. In section 3 we will ex plain th e ECN m echanism briefly. Section 4 will present o ur sim ulation d esign, p arameter settings, and si mulation resul ts. Fi nally, we conclude our paper in section 5.
2. Active and Passive Queue management From t he poi nt of dropping packets, queue management can be classified into tw o categories; Passive Queue Management (PQM ) and Act ive Queue M anagement (AQM). Drop Tail is a rep resentative PQM alg orithm which only sets a m aximum length for each queue at the router. W hen t he queue l ength i s sm aller t han the
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org maximum length, all packets are accepted, and if the queue reaches its m aximum length all subsequent incoming packets are dropped until queue length decreases to be l ess t han t he maximum l ength. It was shown i n [6] that under heavy load conditi ons, Drop Tail routers cause global sy nchronization, a phenom enon in which all senders sharing the same bottleneck router/link shut down their transmission windows at almost the same time. To sol ve t his probl em, AQM was proposed. AQM techniques try to detect and r eact to congestion before its consequences such as queuing delay and packet loss. One well known AQM al gorithm i s random early detection (RED) [2] whi ch has been recom mended by the IETF as the defaul t AQM schem e for routers of next generation networks (NGN) [4, 5] . The basic idea of RED is that a router det ects congest ion earl y by computing the average queue l ength avg, and set s t wo buffer t hresholds; maximum threshold (maxth) and minimum threshold (minth) for packet drop. The average queue length at tim e t, is computed using an exponent ially wei ghted m oving average (EWMA) of t he previous queue length, as shown in equation 1: avg (t ) (1 W q ) avg (t 1) W q Q (t ) (1) where Q i s t he current queue l ength and W q is a weight parameter, 0 Wq 1. Wq det ermines how rapidly avg changes in response to changes in Q. The t wo t hresholds are used t o est ablish t hree zones. When the average queue length is sm aller th an th e minth, RED is in the normal operat ion zone and al l packet s are accepted. On the other hand, if the average queue length is larger than th e m axth, R ED i s i n t he congest ion cont rol region and all the arriving p ackets will be dropped. If the average queue length is between both thresholds, RED is in t he congest ion avoi dance regi on and t he packet s are discarded with a g iven d rop p robability. The probability that a packet arriving at t he R ED queue i s dropped depends on t he average queue l ength, t he num ber of packets si nce l ast di scard, and t he maximum drop probability p arameter m axp. Th e d rop p robability Pa is computed as: Pb (2) Pa (1 count . Pb ) where count is the num ber of accepted packets since the last packet discard and Pb is a temporary probability, given by: Pb max p * ( avg - min th ) /( max th - min th ) (3) whose maximal value given by maxp is reached when the average queue length i s equal t o m axth. For a const ant average queue l ength, all incoming packets have the same probability to get dropped. As a result, RED drops packets in proportion to the connections’ share of the bandwidth. The RED algorithm is shown in Fig 1.
14
for each packet arrival compute the average queue size avg if minth avg < maxth calculate probability Pa drop packet with probability Pa else if maxth avg drop the arriving packet Fig. 1 RED algorithm.
3. Explicit Congestion Notification (ECN) ECN [7] is a bi t-based end-t o-end congest ion avoi dance mechanism proposed for TC P/IP networks and incorporated into RFC 2481. R ED can be easi ly extended to support EC N. EC N ext ends R ED enabled routers to mark packets instead of dropping t hem when t he avg i s between minth and m axth. This mechanism turns out to be useful to protocols lik e TCP th at are sen sitive to ev en a single packet drop. Upon receipt of m arked packets, the TCP receiver informs the sender (in the subsequent ACK) about incipient congestion. Th is will in turn trigger the congestion avoidance al gorithm once per wi ndow. EC N requires support from both the routers and the end hosts. With ECN traffic, ro uters h ave to id entify wh ether a packet is ECN capable or not. R outers m ark onl y EC N capable packets when congestion occurs. This requirement has resu lted in allo tting two b its in th e IP header. The ECN Capable Transport (ECT) bit is set by the sender end system if both the end system s are ECN capable. This is confirmed during the TC P’s pre-negot iation connect ion setup phase. On congest ion, t he rout er m arks the ECT packets usi ng t he C ongestion Experi enced (CE) bit on their way to the receiver, with a probability proportional to the average queue size as done in RED [8].
4. Simulation Results The si mulations present ed i n t his paper were performed with the OPNET [9 ] network simulation tool. To enable a fair comparison of our work with previous works, we have used the simulation configuration shown i n Fig. 2, whi ch has been used by previous researchers [1, 8, 10]. There are two routers connected by a DS1 link and all other links are 10BaseT. Five clients run FTP sessi ons to the five servers (chosen randomly) across t he rout ers. The l ink bet ween router A and rout er B i s m ade a l ow bandwi dth l ink t o create enough congest ion t o anal yze t he effect of Drop Tail, RED an d RED-with -ECN alg orithms in router A. The RED parameters were set as follows, as recommended in the original paper [2], minth = 25, maxth = 75, W q = 1/2 9 and maxp = 0.1. The maximum queue size at the congested
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org router was set to 100 packets. Simulations were ran for 10 minutes.
Fig. 2 Simulation topology.
In this experiment we will com pare two popular queue management algorithms, RED and Drop Tai l, in terms of queue size, queuing delay, queuing delay variations, packet drop rate and bandwidth utilization. We will also study the effect of buffer behavior on QoS m easurements as well as considering t he effect s of usi ng EC N on RED algorithm Performance. Furthermore, we try to find the difference of utilization o f ro uter p rocessor (CPU u tilization) b etween active and passive queue management algorithms. Figure 3 in dicates th e b uffer u tilization o f th e q ueue management m echanisms versus si mulation t ime (buffer behavior).
Fig. 3 Buffer usage vs. simulation time.
To st udy t he effect s of buffer behavior on QoS measurements, such as del ay and droppi ng packet s, we predict each one of QoS measurement according to Fig. 3, and then compare our predi ctions and act ual si mulation results.
15
4.1 Comparing the queuing delay As it is observed from Fig. 3, Drop Tai l has the maximum buffer usage duri ng t he si mulation period related to the other queue m anagement algorithm s. This is because of the reason that in this m ethod, new packets are accepted until there was still a p acket to entrance to the queue and queue didn' t reach its m aximum length. Hence we expected that th is alg orithm will h ave th e m aximum queuing delay. The averages queuing delay of these three algorithms were presented in Fig. 4. According t o t his fi gure, as we have expected, RED al gorithm due t o i ts random packet dropping mechanism, has the minimum queuing delay and Drop Tail algorithm due to its high buffer utilization, has the maximum queuing delay. As it is observed from Fi g. 3 and Fi g. 4 usi ng t he EC N mechanism will in crease th e b uffer u tilization an d consequently increasing the queuing del ay. The reason i s that, as it was said before, when the average queue length exceeds the minimum threshold, RED-with-ECN instead of dropping packet, set a specific bit in the header of t hat packet and forward it to th e receiver. Upon receiving the packet, the receiver will set another bit in its next ACK and n otify th e sen der to red uce its transmission rate. According to Fi g. 4 by i ncreasing t he si mulation t ime, queuing delay for both R ED and R ED-with-ECN are going to be in close. It is because of the fact that by increasing the simulation time, TCP senders receive ACK massages and reduce their transmission rates.
Fig. 4 Average queuing delay.
4.2 Comparing the queuing delay variation As it is shown in Fig. 3 Drop Tail schem e, because of its global sy nchronization probl em, has m ore vari ation in buffer utilization in co mparison with RED alg orithm. In this algorithm queue reaches its m aximum length at 156
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org
16
sec, and buffer was full until 192 second. Then by finishing the waiting time for receiving the ACK messages, without receiving them (becau se of packet dropping in congested rout er), al l TC P senders reduced their transmission rates in th e same time (global synchronization problem). So we est imate t hat Drop Tai l algorithm has more delay variation (jitter) in compression with RED. Figure 5 indicates the accuracy of our estim ate by representing average delay variation. It is observed from this figure that queuing delay variation of RED-with-ECN, which was m ore than RED at commence of the sim ulation, by increasing the tim e and receiving ACK m essages to TC P senders, becam e almost equal to RED. Fig. 6 Packet drop rate.
4.4 Comparing the bandwidth utilization Figure 7 sh ows th e u tilization o f o utgoing lin k, b y router A versus simulation time. C omparing Fi g. 3 and Fi g. 7 indicate that when the RED' s buffer usage is reduced (due to its random packet dropping mechanism), its bandwidth utilization is reduced too. In ad dition, u sing ECN mechanism, which caused the increasing in RED' s buffer utilization, also caused the increasing in RED's b andwidth utilization.
Fig. 5 Average queuing delay variation (jitter).
4.3 Comparing the packet dropping rate Figure 6 indicates packet drop rate for each queue management algorithm versus simulation time. Comparing Fig. 3 and Fi g. 6 indicates that at the same time which the RED's buffer utilization exceed ed the m inimum threshold (25 packets) in Fig. 3 (around second m inute of t he simulation time), this alg orithm started to d rop so me arriving packet s random ly i n Fi g. 6. It cause t hat some TCP senders reduced their transmission rates and therefore RED has l ess packet droppi ng duri ng the simulation period related to the Drop Tail algorithm. Comparing t he packet drop rat e i n R ED and RED-withECN algorithms in Fig. 6 indicates that RED's packet drop rate reduced due to ECN m echanism. Using ECN mechanism in RED is useful for traffics which are sensitive to packet dropping.
Fig. 7 Link utilization.
4.5 Comparing the router CPU utilization Figure 8 shows the percentage of CPU utilization for each queue m anagement al gorithm versus simulation time in router A. Comparing the Fig. 3, Fi g. 6 and Fi g. 8 i ndicate that at the second minute of the simulation time, which the average queue length exceeded the minth (according to Fig. 3) and the RED algorithm st arted t o drop som e arri ving
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org packets randomly (according to Fig. 6), CPU utilization of this alg orithm was in creased related to Drop Tail algorithm in Fig. 7. Also at 156 sec. whi ch Drop Tai l algorithm reaches to its m aximum length and started to packet dropping, its CPU utilization was increased. So we can say queue m anagement al gorithms use t he processor more at t he packet droppi ng m oment. It i s observed that after the time that all algorithms started to packet dropping, their CPU utilization were almost as same as each other. So it can be concl uded that the difference between active and passi ve queue m anagement al gorithms, i n term of CPU u tilization, is m ostly d epend o n th eir difference moment st arting t he packet droppi ng. Packet dropping start moment in different queue management algorithms is depend on buffer si ze and net work t raffic l oad. If the buffer size was sm all o r th e n etwork traffic lo ad was heavy, buffer usage graph woul d have t he more gradient. So a few seconds after queue length exceeds the m inth (for RED algorithm), it will reach to the maximum length for Drop Tai l al gorithm. Thus t he di fference of packet dropping start moment i n R ED and Drop Tai l al gorithms and consequent ly di fference bet ween t hem in CPU utilization, will be decreased.
Fig. 8 CPU utilization (%).
5. Conclusion and results In this paper we presented a simulation-based performance evaluation and com pared Drop Tail queue management algorithm, whi ch used i n t oday's IP net works, and RED queue m anagement al gorithm, whi ch is recommended as the default active queue m anagement scheme for NGN, in different aspect s. The si mulation resul ts show RED outperforms Drop Tail in terms of queuing delay, queuing delay vari ations, and packet drop rat e. W e also noticed that usi ng EC N m echanism has great ly i mproved R ED's packet d rop rate an d b andwidth u tilization performance
17
measures. Queui ng del ay and queui ng del ay variation increasing due to using ECN, will be decreased gradually as the time of si mulation i s progressed and AC K packet received to the senders. This mechanism turns out to be useful to protocols lik e TCP th at are sen sitive to ev en a single packet drop. Al so i t was shown t hat by experimenting buffer’s behavi or, we can have respective comparisons about delay, delay variation, packet drop rate, and bandwidth utilization Qo S m easurements. Fin ally we observed t hat t he di fference bet ween rout er CPU utilization by active and p assive q ueue m anagement algorithms, i s m ostly depend on t heir di fferent moment starting t he packet droppi ng and consequent ly i s depend on the buffer size and the network traffic load.
References
[1] T. Reddy, A. Ahammed, and R. banu, “Performance Comparison of Active Queue M anagement T echniques”, IJCSNS International Journal of Com puter Science and Networ k Secur ity, Vol. 9 No. 2, February 2009, pp. 405-408. [2] S. Floy d, V. Jacobson, “Ra ndom early detection gateway s for congestion avoidance”, IEEE/ACM Transactions on Networking1, 1993, pp. 397-413. [3] A. Bitor ika, M . Robin and M . Huggar d, “A com parative study of Active Queue Managem ent schem es”, In Proceedings International Conference on Communications (ICC 2004), June 2004. [4] D. Que, Z . Chen, B. Chen, “An im provement algor ithm based on RED and its per formance analy sis”, 9th Int. Conf. on Signal Processing, Oct. 2008, pp. 2005- 2008. [5] B. Zheng, M. Atiquzzam an, “ DSRED: A New Queue Management Scheme for the Next Gener ation Networ ks” I EICE T rans. on Communications, Vol. E89-B, No. 3, 2006, pp. 764-774. [6] K. Z hou, K. Yeung, V. L i, “Nonlinear RE D: a simple yet efficient active queue m anagement schem e” Com puter Networ ks: T he Int. Journal of Com puter and Telecommunications Networking, Vol. 50, No. 18, 2006, pp. 3784-3794. [7] K. Ram akrishnan and S. Floy d, “A pr oposal to add Explicit Congestion Notification (ECN) to IP”, Available as RFC 2481. [8] D. Packer , H. Pey ravi “Sim ulation M odeling and Performance Comparison of RE D Var iant Algor ithms,” OPNE T W ORK 2001, Washington DC, Aug 27-31, 2001. [9] OPNET Technologies, Inc. http://www.opnet.com. [10] W. Eddy, M. Allm an, “A com parison of RE D's by te and packet modes,” Computer Networ ks, Vol. 42, No. 2, June 2003, pp. 261280.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
23
A Minimum-Process Coordinated Checkpointing Protocol For Mobile Distributed System Praveen Kumar1 and Ajay Khunteta2 1 Department of Computer Science & Engineering Meerut Institute of Engineering & Technology, Meerut, India, Pin-125005 2 Singhaniya University Pechri, Rajasthan, India
Abstract
While dealing with Mobile Distributed systems, we come across some issues like: m obility, low bandwidth of wireless channels and lack of stable storage on mobile nodes, disconnections, limited battery power and high failu re rate of mobile nodes. These issues make traditional checkpointing techniques designed for Distributed sy stems unsuitabl e for Mobile environments. In this paper, we design a minimum process algorithm for Mobile Distributed systems, where no us eless checkpoints are taken and an effort has been made to op timize the blocking of processes. We propose to delay the processi ng of selective messages at the receiver end only during the chec kpointing period. A Process is allowed to perform its normal computations a nd send messages during its blocking period. In this way , we try to keep blocking of processes to bare m inimum. W e captured the transitive dependencies during the normal execution by piggybacking dependency vectors onto computati onal messages. In this way , we try to reduce the Checkpointing time by avoiding formation of Checkpointing tree. The Z-dependencies are well taken care of. The proposed scheme forces zero useless checkpoints at the cost of very small blocking.
1. Introduction Checkpoint is defined as a designated place in a program at which normal p rocess is in terrupted sp ecifically to preserve the status inform ation necessary to allow resumption of processing at a later time. A checkpoint is a local state of a process sa ved on st able st orage. B y periodically i nvoking t he checkpoi nting process, one can save the status of a program at regular intervals [3], [4]. If there is a failu re, o ne m ay restart computation from the last checkpoints, t hereby, avoi ding repeat ing computation from t he begi nning. The process of resuming computation by rolling back t o a saved st ate i s called rollback recovery [6]. In a distributed system , since
the processes in the system do not share m emory, a global state of the system is defined as a set of local states, one from each process. The state of channels corresponding to a global state is t he set of m essages sent but not y et received [7]. A message whose receive event is recorded, but its send event is lo st. A g lobal state is said to be “consistent” if it contains no orphan message. To recover from a failure, the system restarts its execution from a previ ous consi stent global st ate saved on t he stable storage during fault-free execution. In di stributed sy stems, checkpointing can be independent, coordi nated [3] , [8] , [11], [15] or quasisynchronous [2] , [9] . M essage Loggi ng i s al so used for fault tolerance in di stributed sy stems [7] , [14] . Under t he asynchronous approach, chec kpoints at each process are taken i ndependently wi thout any sy nchronization am ong the processes. Because of absence of synchronization, there is no guarant ee that a set of l ocal checkpoints taken will be a consistent set of checkpoints. It may require cascaded rollbacks that may lead to the initial state d ue to domino-effect [7]. In coordi nated or sy nchronous Checkpointing, processes take checkpoints in such a manner that the resulting global state is consistent. Mo stly it fo llows two -phase co mmit structure [3] , [8] , [11] , [22] . In the first phase, processes take t entative checkpoi nts and i n t he second phase, these are made permanent. The m ain advantage is that only one permanent checkpoint and at m ost one t entative checkpoint is required to be st ored. In t he case of a faul t, processes rollback to the last checkpointed state. It avoids the dom ino-effect wi thout requi ring al l checkpoints t o be coordi nated [2] , [7] , [9]. In these protocols, processes take two ki nds of checkpoi nts, l ocal and forced. Local checkpoi nts can be t aken independently, whi le forced checkpoi nts are t aken to guarantee the eventual progress of the recovery line and to
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org minimize usel ess checkpoi nts. Pj is di rectly dependent upon Pk only if th ere exists m such that Pj receives m from Pk in the current CI and Pk has not t aken i ts perm anent checkpoint aft er sendi ng m. A process P i is in the minimum set only if checkpoint initiator process is transitively dependent upon it. In minimum-process coordinated checkpoi nting al gorithms, only a subset of interacting processes (called minimum set) are requi red to take checkpoints in an initiation. The Chandy-Lamport [6] algorithm i s t he earl iest nonblocking all-process coordinated checkpointing algorithm. In th is alg orithm, markers are sent al ong al l channel s i n the network which lead s to a m essage co mplexity o f O(N2), and requi res channels to be FIFO. Elnozahy et al. [8] proposed an al l-process non-bl ocking sy nchronous checkpointing al gorithm wi th a m essage com plexity of O(N). In coordi nated checkpoi nting prot ocols, we m ay require piggybacking of i nteger csn (checkpoint sequence number) on normal messages [5], [8], [13], [19], [22]. The exi stence of m obile nodes i n a distributed system introduces new issues that need proper handl ing whi le designing a checkpoi nting al gorithm for such systems. These issues are m obility, di sconnection, finite power source, vulnerable to physical dam age, l ack of st able storage et c. These i ssues make traditional checkpointing techniques unsui table t o checkpoi nt m obile di stributed systems [1], [5], [15]. To take a checkpoint, an MH has to transfer a large amount of checkpoint data to its local MSS over the wireless network. Since the wireless network has low bandwi dth and M Hs have l ow com putation power, all-process checkpointing will waste the scarce resources of the mobile sy stem on every checkpoi nt. Prakash and Singhal [15] gave minimum-process coordinated checkpointing protocol for mobile distributed systems. A good checkpoi nting prot ocol for m obile di stributed systems should have low overheads on M Hs and wi reless channels and shoul d avoi d awakeni ng of M Hs i n doze mode operat ion. The di sconnection of M Hs shoul d not lead t o i nfinite wai t st ate. The al gorithm shoul d be nonintrusive and shoul d force m inimum number of processes to t ake t heir l ocal checkpoi nts [15] . In m inimum-process coordinated checkpoi nting al gorithms, som e blocking of the processes takes place [4], [11], or some useless checkpoints are taken [5], [13], [19]. Cao and Si nghal [5] achi eved non-i ntrusiveness i n the minimum-process algorithm by introducing the concept of m utable checkpoi nts. The num ber of useless checkpoints in [5] may be exceedingly high in som e situations [19] . Kum ar et . al [19] and Kum ar et. al [13] reduced t he hei ght of t he checkpoi nting t ree and t he number of usel ess checkpoi nts by keepi ng nonintrusiveness in tact, at th e ex tra co st o f maintaining and collecting dependency vect ors, com puting t he minimum
24
set and broadcast ing the same on t he static network along with the checkpoint request. Koo and Toeg [11], and C ao and Si nghal [4] proposed minimum-process bl ocking coordi nated checkpointing algorithms. Neves et al. [12] gave a l oosely synchronized coordinated prot ocol t hat rem oves t he overhead of synchronization. Higaki and Taki zawa [10] proposed a hybrid checkpoi nting prot ocol where t he m obile st ations take checkpoi nts asy nchronously and fixed ones synchronously. Kumar and Kum ar [29] proposed a minimum-process coordi nated checkpoi nting al gorithm where the number of useless checkpoints and blocking are reduced by using a probabilistic approach. A process takes its mutable checkpoint only if the probability that it will get the checkpoint request in the current initiation is high. To bal ance t he checkpoi nting overhead and t he l oss of computation on recovery , P Kum ar [24] proposed a hybrid-coordinated checkpointing prot ocol for m obile distributed systems, where an al l-process checkpoi nt i s taken aft er execut ing m inimum-process checkpoi nting algorithm for a certain number of times. Transferring the checkpoint of an M H t o i ts l ocal M SS may have a l arge overhead i n t erms of battery consumption an d ch annel u tilization. To reduce such an overhead, an i ncremental checkpoi nting technique could be used [16] . Onl y t he information, which changed since last checkpoint, is transferred to the MSS. In t he present st udy, we purpose a minimum process coordinated checkpoi nting al gorithm for M obile Distributed Systems in which no usel ess checkpoi nts are taken and the blocking of processes i s reduced t o bare minimum.
2. System Model We u se th e system m odel p resented in [2], [4]. In this model, a mobile computing system consi sts of n m obile hosts (M Hs), and m m obile support stations (MSSs), where n > m . A cel l is a l ogical or geographi cal coverage area under an MSS. An M H can di rectly com municate with an MSS M i only if it is p resent in the cell serviced by Mi. At any time, an M H belongs to only one cel l or may be disconnected. The st atic net work provi des rel iable First-In-First-Out (FIFO) delivery of m essages between any two MSSs with arbitrary message laten cy. Sim ilarly, the wireless n etwork with in a cell en sures reliable FIFO delivery of messages between an MSS and an MH. In this paper, we consi der a di stributed com putation i n a mobile com puting sy stem t hat consi sts of N processes, running concurrently on different M Hs or M SSs. For simplicity, we assum e that each MH runs one process. Message passing is t he onl y way of com munication. The computation is asynchronous. The processes do not share memory or cl ock. Each process progresses at its own
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org speed and m essages are exchanged through reliable channels, whose transmission del ays are fi nite but arbitrary. A process in the cell of MSS m eans the process is either running on the MSS or on an MH supported by it. It al so i ncludes t he processes of M Hs, which have been disconnected from t he M SS but t heir checkpoi nt rel ated information is still with this MSS. We also assume that the processes are non-deterministic. The i th CI (checkpointing interval) of a process denot es al l t he computation performed b etween its ith an d (i+1 )th checkpoi nt, including the ith checkpoint but not the (i+1)th checkpoint.
3. Basic Idea During t he execut ion of checkpoi nting al gorithm, a process P i may receive m from Pj such that P j has t aken its tentative checkpoint for the current initiation whereas P i has not taken. If Pi processes m and it receives checkpoint request later on and takes its checkpoint, then m will become orphan i n t he recorded gl obal state. We propose that such m essages should be buffered at the receiver end. In t he present di scussion, P i processes m onl y aft er taking i ts t entative checkpoi nt i f i t i s a m ember of the minimum s et; o therwise, P i processes m aft er get ting t he exact minimum set and knowing that it is not a member of the minimum set. P1 __________________________________ R1[0001] t1 P2 __________________________________ R2[0010] m3 {after m3, R3 is 1101} m2[1100] P3 __________________________________ R3[0100] m1[1000] P4 __________________________________ R4[1000] Time Fig. 3.1 Basic Idea In the fi gure 3.1 P 4 sends m 1 to P 3 al ong wi th i ts own dependency vect or R4[1000]. W hen P 3 receives m 1 it updates its own dependency vect or by t aking l ogical OR of R 4 & R 3[0100], which comes out to be 1100. When P3 send m 2 to P 2, it appends R 3[1100] along wi th m 2. W hen P2 receive m 2, it updates its own dependency vector R 2 by taking l ogical OR of R 2 and R 3, whi ch com es out t o be [1110]. In this way, partial transitive dependencies are captured during normal computation. It shoul d be not ed that all the transitive dependencies are not captured during normal computation. At time t1, the dependency vector of
25
P2 shows that P2 is not transitively dependent upon P 1, due to m3 and m2.
3.1 Example We explain our al gorithm with an exam ple. P1, P 2, P 3, P 4 and P 5 are processes with initial dependency set [00001], [00010], [00100], [01000] and [10000], respectively. P1___________________________________________ m1, [00001] R1[00001] P2___________________________________________ R2[00010] t2 m2, [00011] t1 P3___________________________________________ R3[00100] m3 [01000] m4 P4___________________________________________ R4[01000] P5___________________________________________ R5[10000] Fig 3.2 An Example { indicate message, indicate request of checkpoint, Ri represent the set of dependency.} At time t 1, P3 initiates checkpointing with dependency set [00111], therefore it sends the checkpointing request to P1 and P 2 only, which in tu rn tak es th eir ten tative checkpoints. Aft er t aking i ts t entative checkpoi nting, P 3 sends m 4 to P 4. W hen P 4 receives m 4, its fin d th at P 3 has taken its t entative checkpoi nt before sendi ng m 4 because CSN (checkpoint sequence num ber) of P 3 is 1 at tim e o f sending m 4; therefore, P4 buffers m 4. W hen P 2 tak es its tentative checkpoi nt, i t fi nd t hat i t i s dependent upon P 4 due to m3 and P 4 is not in the minimum set of dependency computed so far; therefore, P2 send checkpoi nt request to P4. After taking its tentative checkpoint, P4 process m 4. At time t 2, P 3 receives response from all processes and sends commit request to all processes al ong with exact minimal set of dependency , whi ch i s not shown in the figure. Hence, the m essages, which can becom e orphan, are buffered at the receiver end. A process processes the buffered m essages onl y aft er taking its tentative checkpoint or after getting the commit request.
4. Data Structures Here, we describe the data structures used in the proposed checkpointing protocol. A pr ocess on MH that initiates checkpointing, is called initiator process and its local MSS is called initiator MSS. If th e in itiator p rocess is o n an
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org MSS, then th e MSS is th e in itiator MSS. All d ata structures are initialized on completion of a checkpointing process, if not mentioned explicitly. Pr_csni: A m onotonically i ncreasing integer checkpoint sequence number for each process. It is increm ented by 1 on tentative checkpoint. td_vecti []: It is a bit array of length n for n process i n the system. td_vect i[j] = 1 im plies P i is tran sitively d ependent upon P j. W hen P i receives m from P j such that P j has not taken any perm anent checkpoi nt aft er sendi ng m then Pi sets td _vecti[j]= 1. W hen P i commit i ts checkpoi nt, i t set s td_vecti[] = 0 for all processes except for itself which is initialized to 1. chkpt-sti: A boolean whi ch i s set t o ‘1’ when P i tak es a tentative checkpoint; on commit or abort, it is reset to zero m_vect[]: A bi t array of si ze n for n processes in the systems. W hen P i starts checkpoi nting procedures, i t computes tentative m inimum set as fo llows: m _vect[j] = td_vecti[j] where j=1, 2, …., n. TC[]: An array of si ze n t o save i nformation about t he processes which have t aken t heir t entative checkpoi nts. When process P j t akes i ts t entative checkpoi nt then jth bit of this vector is set to 1. It is initialized to all zeros in the beginning of t he checkpoi nting process. It i s maintained by the checkpoint initiator MSS only. Max_time: i t i s a fl ag used t o provide timing in checkpointing operation. It is initialized to zero when timer is set and becom es ‘1’ when maximum allowable time for collecting global checkpoint expires. MSS_plist[]: A bit array of length n for n processes which is maintained at each MSS MSS_plist K[j] =1 im plies each process Pj is running on MSSk. If Pj is disconnected, then it checkpoint related information is on MSSk. MSS_chk_taken: A bit array of length n bits maintained by the MSS. MSS_chk_taken [j]=1 implies P j which is in the cell of MSS has taken its tentative checkpoint. MSS_chk_request: A bit array of length n at each MSS. The j th b it o f th is array is set to ‘1 ’ whenever initiator sends t he checkpoi nt request t o P j and P j is in th e cell o f this MSS. MSS_fail_bit: A flag maintained on every MSS, initialized to ‘0’; set to ‘1’ when any process in the cell of MSS fails to take tentative checkpoint. Pin: The process which has initiated the checkpointing operation. MSSin: The MSS, which has Pin in its cell. p_csnin: checkpoint sequence number of initiator process. g_chkpt: A fl ag whi ch i ndicates t hat som e global checkpoint is being saved. csn[]: An array of size n, m aintained on every MSS, for n processes. csn[i] represents the m ost recently com mitted checkpoint sequence num ber of Pi. After th e co mmit operation, if m _vect[i] = 1 th en csn [i] is in cremented. It should be not ed that entries in this array are updat ed only
26
after converting t entative checkpoi nts i n t o perm anent checkpoints and not after taking tentative checkpoints. m_vect1[]: An array of size n m aintained on every MSS. It contains those new processes which are found on getting checkpoint request from initiator. m_vect2 []: An array of size n. for all j such that m _vect1 [j] 0, m_vect2= m_vect2 m_vect1. m_vect3[]: An array of length n; on receiving m _vect3[], m_vect[], m _vect1[] al ong wi th checkpoi nt request [c_req] or on t he com putation of m_vect1[] locally: m_vect3[]=m_vect3[] c_req.m_vect3[]; m_vect3[]=m_vect3[]m_vect[]; m_vect3[]=m_vect3[]c_req.m_vect1[]; m_vect3[]=m_vect3[] m_vect1[]; m_vect3[] m aintains t he best l ocal knowl edge of the minimum set at an MSS.
4.1 Computation of m_vect[], m_vect1[], m_vect2[], m_vect3[]: 1. Suppose a process Pr wants to initiate checkpointing procedure. Its send its request to its local MSS, say MSSr.. MSSr m aintains t he dependency vect or of P r (say td_vectr[]). M SSr coordi nates checkpoi nting on behal f of Pr. It computes tentative minimum set as follows:
i=1,n m_vect[i] = td_vectr[i]
2. On receiving m_vect[] from MSSr, any MSS (say MSSS) computes the m_vect1[] as follows: Suppose M SSs m aintains the process Pj such that P j MSSs and Pj m_vect m_vect1[i]=1 iff m_vect[i]=0 and td_vectj[i]=1 m_vect1[] m aintains t he new processes found for t he minimum set when a process receives the checkpoint request. m_vect2=m_vect2 U m_vect1 i, m_vect1[i]=0 3. m_vect3= m_vect U m_vect2 MSSin sends c_req t o MSSs along with m_vect[]and some process (say P k) i s found at M SSs, which tak es th e checkpoint t o t his c_req. Al l M SSs m aintains t he processes of minimum set to the best of their knowledge in m_vect3. It is required t o m inimize dupl icate checkpoi nt requests. Suppose, t here exists some process (say P l) such that P k i s di rectly dependent upon P l and P l is n ot in th e m_vect3, then MSSs sends c_req t o P l. The new processes found for t he m inimum set whi le execut ing a pot ential checkpoint request at an M SS are st ored i n m _vect1. When an MSS finds that all the local processes, which were asked t o t ake checkpoi nts, have t aken t heir checkpoints, it sends the response to the MSSin along with m_vect2; so that MSS in may updat e i ts knowl edge about
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org minimum set and wai t for t he new processes before sending commit. In this way, MSS in sends commit only if all th e p rocesses in th e m inimum set h ave taken their tentative checkpoints.
5. The Checkpointing Protocol As th e wireless b andwidth is a scarce co mmodity in mobile systems; therefore; we impose minimum burdon on wireless channels. The local MSS of an MH acts on behalf of the process running on MH. We pi ggyback checkpoi nt sequence num bers and dependency vectors ont o norm al com putation m essages, but this information is not sent on wi reless channels. The local MSS of an MH, strip s all th e additional information from t he com putation m essage and sends i t to the concerned M H. The dependency vect or of a process running on an MH is maintained by its local MSS. Our algorithm is distributed in nature in the sense that any process can initiate checkpointing. If two processes initiate checkpointing concurrently, t hen t he checkpoi nt i mitator of the lower p rocess ID will p revail. Th e lo cal MSS o f a process coordi nates checkpoi nting on i ts behalf. Suppose two processes P i and P j st arts checkpoi nting concurrent ly and MSSp and MSSq are their local MSS respectively then MSSp and M SSq will send checkpoint requests along with tentative minimum set to all the MSS’s. MSSp will receive the checkpoi nt request of MMSq and M MSq will receive the checkpoint request of MSSp. Suppose Process-ID of P i is less th an Process-ID of P j, then the checkpoint initiates of P i will p revail. An y o ther MSS will automatically ignore the request of P j because every M SS will com pare the process id of Pi and Pj. We propose that any process in the system can initiate the checkpointing operat ion. W hen a process Pin starts checkpointing procedure, i t send i ts request t o its local MSS say MSS in. MSS in com putes the tentative m inimum set m_vect[] as follows: i=1,n m_vect[i] = td_vect[i] MSSin coordinates checkpointing process on behalf of Pin. We want to em phasize that td_vect in[] cont ains t he processes on whi ch P in transitively d epends an d th e set is not complete. MSSin sends c-req to al l M SS’s al ong wi th m _vectin[]. When an MSS say MSS p receives c-req; it sends the c-req to all such process which are running in it and are also the member o f m_ vectin[]. Suppose P j gets t he checkpoi nt request at MSSp Now we find any process Pk such that P k does not belong to m_vectin[] and P k belongs to td_vectj[]. In this case, P k is also in cluded in th e m inimum set. During checkpointing suppose P i tak es it ten tative checkpoint and aft er t hat i t send m t o P j such that P j has not taken it tentative checkpoint at the time of receiving m.
27
If P j receive m and it gets chec kpoint request later on then m will become orphan. In o rder to h andle th is situ ation, we buffer m at P j. P j receive m after taking its tentative checkpoint i f i t i s m ember of minimum set; otherwise it process m on commit. For a disconnected MH that is a m ember of minimum set, the MSS that has i ts disconnected checkpoint, converts its disconnected checkpoint into tentative one. W hen a MSS learns t hat i ts concerned processes i n its cell have taken their tentative checkpoints, it sends the response to MSSin. On receiving positive response from all concerned MSSs, the MSS in issu es th e co mmit req uest to all MSSs. On commit when a process learns t hat i t has buffered some message and has not received the form al tentative checkpointing request from any process, t hen it processes the buffered messages.
5.1 Formal Outline of the checkpointing Algorithm: 5.1.1 Actions taken when Pi sends m to Pj: send(Pi, Pj, m, pr_csni,td_vecti[]); //Pi piggybacks its o wn csn an d tran sitive dependency vector onto m. 5.1.2 Algorithm executed at initiator MSS (say MSSin) Suppose P in initiates checkpointing. P in sends t he request to MSSin. MSSin computes m_vect [Refer section 4.1]. (1)On t he basi s of com puted m _vect, M SSin com putes m_vect1, m_vect2, m_vect3 [Refer section 4.1]. (2) m_vect = m_vect3. (3) MSSin sends c_req to all MSSs along-with m_vect[]. (4) Set max-time. (5) Wait for response. (6) On receiving response (Pin, MSSin, MSSs, mss_ chk_taken, m_vect2, mss_fail_bit) or at max_time (a) If (m ax_time)OR(mss_fail_bit){ send m essage abort (P in, MSS in, pr_csn in} to all MSS s, Ex it; //Maximum allocated time expired or some process failed to take checkpoint (b) m_vect[] = m _vect[]U m _vect2[]. [“U” is a set union operator] (c) TC[] = TC[] U mss_chk_taken[] (7) For (k=0;kcsn[j])); {rec (m); td_vectj[i]= 1} //P i & P j, bot h, have t aken their tentative checkpoints Case 4: If(( m.pr_csni>csn[i]) AND (pr_csnj=csn[j])); {Pj buffers m } Pi h as tak en its ten tative checkpoint // before sending m while Pj has not. 5.1.4 Algorithm executed at any MSS (say MSSs) (1) Wait for Response (2) Upon receiving m essage c_req (P in, MSS in, p_csn i, m_vect) from MSSin (i)For any Pi such that mss_plists[i] =1 m_vect[i]=1; send c_req to Pi (ii) ++pr_csni; mss_chk_request[i]=1, chkpt_sti=1 (iii)Compute m_vect1, m_vect2, m_vect3 //Refer Section 4.1 (iv) If i such that m_vect1[i]=1; send c_req to Pi. //m_vect1 contains the new processes found for the //minimum set (3) On receiving c_req from some other MSS say MSSp i such that(( mssp. m_vect1[i] = 1) (mss_p_mss[i]= 1) (mss_chk_req=1)) { send c_req to Pi; compute m_vect1, m_vect2, m_vect3} If j such that m_vect1[j]=1; send c_req to Pj; i, m_vect1[i]=0; (4) On receiving response to checkpointing from Pj (i) If (P j has taken t he t entative checkpoi nt successfully t he m ss_chk_taken[j]= 1 el se m ss_set fail_bit.) (ii) If (mss_fail_bit) (j mss_chk_taken[j] = mss_chk_request[j]; Send response (Pin, MSSin,msss, mss_chk_taken, mss_fail_bit, m_vect2) to MSSin; (5) On receiving commit(). (i) C onvert t he t entative checkpoi nts in to permanent ones and discard old permanent checkpoints. (ii) Process buffered messages, if any;. (iii) j such that m_vect[j]=1, csn[j]++; (iv) Initialize relevant data structures. (6) On receiving abort(). Discard t he t entative checkpoi nts and induced checkpoints, if any. Update relevant variables. 5.1.5 Algorithm executed at any process Pi; On receiving tentative checkpoint request, Take tentative checkpoint and inform local MSS.
28
6. Handling Node Mobility and Disconnections An M H m ay be di sconnected from t he net work for an arbitrary peri od of t ime. The C heckpointing al gorithm may generate a request for such M H to take a checkpoi nt. Delaying a response m ay si gnificantly i ncrease t he completion t ime of t he checkpoi nting al gorithm. We propose the following solution to deal with disconnections that may lead to infinite wait state. When an MH, say MHi, disconnects from an MSS, say MSSk, MHi takes its own checkpoint, say disconnect_ckpti, and transfers it to MSSk. MSSk stores all th e relevant data structures and disconnect_ckpti of MHi on stable storage. During disconnection period, MSSk acts on behal f of MHi as follows. In m inimum-process checkpoi nting, i f MHi is in th e minset[], disconnect_ckpti is considered as MHi’s checkpoint for the current in itiation. In all-process checkpointing, i f MHi’s disconnect_ckpti is already converted into perm anent one, then the committed checkpoint is considered as t he checkpoint for the current initiation; o therwise, disconnect_ckpti is considered. On global checkpoi nt com mit, MSSk also updat es MHi’s data structures, e.g., ddv[], cci etc. On the receipt of messages for MHi, MSSk does not update MHi’s ddv[] but maintains two message queues, say old_m_q and new_m_q, to store the messages as described below. On the receipt of a message m for MH i at MSS k from any other process: if((m.cci= = ccii (m.cci= =ncii) (matd[j, m.cci]= =1)) add (m, new_m_q); // keep the message in new_m_q else add( m, old_m_q); On all-process checkpoint commit: Merge new_m_q to old_m_q; Free(new_m_q); When MHi, enters in the cell of MSSj, it is connected to the MSSj if g_chkptj is reset. Otherwise, it waits for g_chkptj to be reset . B efore connect ion, MSSj collects MHi’s ddv[], cci, new_m_q, old_m_q from MSSk; and MSSk discards MHi’s support i nformation and disconnect_ckpti. MSSj sends the m essages in old_m_q to MHi wi thout updat ing the ddv[], but m essages in new_m_q, updat e ddv[] of MHi.
6.1 Handling Failures during checkpointing An MH may fail during checkpointing process. If an MH fails after taking i ts t entative checkpoi nt or i f i t i s not a member of m inimum set , t hen t he checkpointing procedure can be com pleted uni nterruptedly. If a process fails duri ng checkpoi nting, t hen our st raight forward approach is to discard the whole checkpointing operation.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org The failed process will not be able to respond to the initiator’s request and the initiator will detect the failure by timeout and will discard the complete checkpointing operation. If the initiator fails after sen ding co mmit, th e checkpointing process can be consi dered complete. If the initiator fails during checkpointi ng, then som e processes, waiting fo r co mmit will tim e o ut an d will issu e ab ort on his own. Kim and Park [17] proposed t hat a process com mits its tentative checkpoints if none of t he processes, on whi ch it transitively depends, fails; and the consistent recovery line is ad vanced fo r th ose p rocesses th at co mmitted th eir checkpoints. The initiator a nd other processes, which transitively d epend o n th e failed process, have to abort their tentative checkpoints. Thus, i n case of a node fai lure during checkpointing, t otal abort of t he checkpoi nting i s avoided.
7. Correctness Proof In this section, we prove t hat our checkpoi nt al gorithm collects a consi stent gl obal checkpoi nting st ate. W e assume that the system is in consistent state when a process initiates checkpointing. Theorem: The gl obal checkpoi nting st ate creat ed by t he ith iteration of the checkpointing protocol is consistent. Proof: Let gl obal_csi ={C1,x, C 2,y,............,Cn,z} be som e consistent global state created by our algorithm, where Ci,x is the xth checkpoint of Pi. The collected global checkpointing state will be inconsistent only if there is a orphan message m sent by Pi to P j such that C i,x and C j,y are in the global state for som e iteration of t he checkpoi nting operat ion. W e prove by contradiction that no such m essage exists. There are following four cases: Case 1: Pi m_vect[] Pj m_vect[] (Pi bel ongs t o t he minimum set and Pj not) As P i has t aken t he perm anent checkpoi nt i n the current initiation an d P j has t aken t he perm anent checkpoi nt i n some previous in itiation; th erefore we can say th at Cix ; (‘ ‘ i s t he Lam port’s happened before Cjy relation); we have already assum ed that rec(m ) Cjy Cix send (m) rec(m) Cjy Cix send(m) rec (m) send (m) Hence it is a contradiction. Case 2: Pi m_vect[] Pj belong to the minimum set)
m_vect[] (Pi and P j bot h
29
Both P i and P j have t aken t heir perm anent checkpoi nts during the current initiation; the following possibilities can take place: Pi sends m aft er commit and Pj receives m before taking the t entative checkpoi nt. As Pj m_vect[], th e in itiator MSS can issue com mit only after P j has taken its ten tative checkpoint and inform the initiator. Therefore rec(m) at P j can not take place before Pj takes its tentative checkpoint. Suppose P i sends m aft er t aking t he t entative checkpoint and P j receive m before taking its tentative checkpoint. In this case, when P j will receive m, it will check the piggybacked P r_csn of P i along with m and will conclude that P i has taken t entative checkpoi nt for t he new initiations and P j has not taken its tentative checkpoint for this in itiation. Th erefore, Pj will p rocess m o nly after P j takes it tentative checkpoint. Hence the receiver of m at Pj can not occur before taking its tentative checkpointing. Case 3: Pi m_vect[] Pj minimum set and Pi not)
m_vect[] ( Pj belongs to the
Checkpoint C ix has been t aken by P i in som e previ ous initiation and checkpoint C jy has been t aken by Pj in th e current in itiation. W hen P j h as tak en its ten tative checkpoint, it will find that P j is dependent upon P i and P i is n ot in th e minimum set co mputed so far. Therefore, P j will sen d th e c_ req to P i and P i will b e included in the minimum set. Hence it is a contradiction. Case 4: Pi m_vect[] Pj m_vect[](Pi and P j both do not belong to the minimum set) In this case, P i and P j will not take checkpoints and therefore no orphan message can exist from Pi to Pj. Hence it is proved that no such orphan message is possible in the recorded global state collected by the proposed algorithm. Hence, the proposed al gorithm leads to the consistent global state.
8. A Performance Evaluation We compare our algorithm with the Koo and Toueg (KT) [11] algorithm, and Cao and Singhal (CS) [4] algorithm on different parameters. (1) In CS algorithm, all processes are bl ocked. In t he KT and the proposed algorithm onl y sel ective processes are blocked. (2) In KT algorithm, a process is blocked, during the time, when it takes its tentative checkpoint and receives com mit or abort from the initiator process. (3) In C S algorithm, a process i s blocked during the time, it sen ds its d ependency v ector to th e initiator MSS and receives checkpoint request al ong with the m inimum set.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org In the proposed prot ocol, a process i s blocked during the period, it receives m of higher CSN and it recues checkpoint request or commit message. In CS alg orithm, in itiator MSS co llects d ependency vectors of all processes, com putes m inimum set and broadcasts minimum set to all MSSs. In KT algorithm and in the proposed protocol, no such step is taken. In KT alg orithm, tran sitive d ependencies are cap tured b y traversing direct dependencies and have a checkpoi nt tree is formed. It may lead to exceedingly high time for global checkpoint collection and the blocking period may also be high. In o ur alg orithm, Tran sitive dependencies are captured duri ng norm al processi ng and hence checkpointing tree is not form ed. Therefore, t he t ime t o collect the global checkpoint w ill be low as com pared to KT algorithm. In CS algorithm, direct dependency vectors are collected in the initiation of the checkpointing algorithm. Therefore, t his al gorithm suffers from hi gh synchronization message overhead. (4) In KT al gorithm and i n t he proposed prot ocol, an integer num ber i s pi ggybacked ont o norm al m essages. In CS algorithm, no such i nformation i s pi ggybacked ont o normal messages. It can not handle the following situation. Pi receives m from P j in the current CI such that Pj has taken some permanent checkpoint after sending m. In this case, Pi does not become causally dependent upon Pj due to receipt of m . In this case, if P i is in th e m inimum set, Pj will unnecessarily be included in the minimum set. (5) Blocking of processes takes place differently in these three protocols as follows. In KT algorithm, processes are not al lowed t o send any m essages. In CS algorithm, processes are not allowed to send or receive any messages. In the proposed prot ocol, a few processes are not allowed to process the selective m essages received only during the checkpointing peri od. A process i s allowed to send messages and perform normal com putations duri ng i ts blocking period. It is even allowed to receive selected messages. (6) We maintain exact dependencies among processes and a best possible knowledge of t he minimum set, computed so far, at the local MSS. In this way, num ber of duplicate checkpoint request s i s reduced as com pared t o t he KT algorithm and no useless checkpoint requests are sent.
8.1 General Comparison with existing nonblocking minimum process algorithms: In th e alg orithms [1 3], [1 9], in itiator process/MSS collects dependency vect ors for al l t he processes and computes t he m inimum set and sends t he checkpointing request to all th e p rocesses with minimum set. These algorithms are non-blocking; the message received during checkpointing may add processes t o t he m inimum set . It
30
suffers from addi tional m essage overhead of sendi ng request to al l processes t o send t heir dependency vect ors and all p rocesses send dependency vectors to the initiator process. B ut i n our al gorithm, no such overhead is imposed. The C ao-Singhal [5] suffers from the formation of checkpoi nting t ree. In our algorithm, theoretically, we can say that the length of the checkpointing tree will be considerably low as compared to algorithm [2], as most of the transitive dependencies are captured during the normal processing. W e do not com pare our al gorithm with Prakash-Singhal [15], as Cao-Singhal proved that there no such algorithm exists [4]. Furthermore, in algorithm [4], transitive dependencies are captured by direct dependencies. Hence the average number of useless chec kpoints requests will be significantly higher t han t he proposed al gorithm. In [5] , huge dat a st ructures are pi ggybacked al ong wi th checkpointing request, because they are unable to maintain exact dependenci es am ong processes. Incorrect dependencies are sol ved by these huge data structures. In our case, no such data structures are piggybacked on checkpointing request and no such usel ess checkpoint requests are sent, because we are able to m aintain exact dependencies among processes and furt hermore, are abl e to cap ture tran sitive d ependencies d uring normal computation at t he cost of pi ggybacking bi t vect or of length n for n processes onto normal computation messages. 8.2 Comparison with other Algorithms: We use following notations to compare our algorithm with other algorithms: Nmss: number of MSSs. Nmh: number of MHs. cost of sending a message from one process to Cpp: another cost of sending a message between any two MSSs. Cst: cost of sending a message from an MH to its local Cwl: MSS (or vice versa). cost of broadcasting a message over static Cbst: network. Csearch: cost incurred to locate an MH and forward a message to its current local MSS, from a source MSS. average message delay in static network. Tst: average message delay in the wireless network. Twl: average delay to save a checkpoint on the stable Tch: storage. It also includes the time to transfer the checkpoint from an MH to its local MSS. N: total number of processes Nmin: number of minimum processes required to take checkpoints. Nmut: number of useless mutable checkpoints [2].
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org Tsearch: average delay incurred to locate an MH and forward a message to its current local MSS. Nucr: average number of useless checkpoint requests in [2]. Ndep: average number of processes on which a process depends. h1: height of the checkpointing tree in Koo-Toueg algorithm [4]. h2: height of the checkpointing tree in the proposed algorithm.: In Koo-Toueg al gorithm [4] and i n the proposed one, t he checkpoint initiator process, say P in sends t he checkpoint request to any process P i if P in is causally dependent upon Pi. Sim ilarly, P i sends the checkpoi nt request t o any process P j if P i is causally dependent upon P j. In this way, a checkpointing tree is formed. Theoretically, we can say that checkpointing tree will not be form ed in our algorithm. B ut due t o Z-dependenci es, a low order checkpointing tree can be form ed, because during norm al computations all th e tran sitive d ependencies are n ot captured. Hence, t he checkpoi nting t ree i n t he proposed scheme will be negligible as co mpared to KT an d CS algorithm in most of the practical situations.
8.3 Performance of our algorithm 8.3.1 The Synchronization message overhead: In the first phase, a process t aking a t entative checkpoi nt needs two system messages: request and repl y. A process may receive m ore than one request for the same checkpoint initiation from differe nt processes. However, we have used som e t echniques t o reduce t he dupl icate checkpoint request s. Thus t he sy stem overhead is approximately 2*N min*Cpp in the first phase. In t he second phase, t he com mit request ed i s broadcast ed on the static network; and the system overhead is Cbst. 8.3.2 Number of processes taking checkpoints: In our algorithm, only minimum number of processes i s required to take their checkpoints.
8.4 A Comparative Study The bl ocking t ime of t he Koo-Toueg [11] prot ocol is highest, fol lowed by C ao-Singhal [4] algorithm. In the algorithms proposed i n [5] , [8] , no bl ocking of processes takes place, but some useless checkpoints are taken, which are discarded on commit. In Elnozahy et al [8] algorithm, all processes t ake checkpoi nts. In t he prot ocols [4] , [11], and the proposed one, onl y m inimum num bers of processes record their checkpoints. The message overhead in the proposed prot ocol is greater than [8], but less than [4], [5] and [11] . In al gorithm [5] , concurrent executions of the algorithm are allo wed, b ut it m ay lead to inconsistencies i n doi ng so [20] . We avoid concurrent
31
executions of the proposed al gorithm. In case, t wo processes concurrently initia te checkpointing, then the
Avg. blocking Time Average No. of checkpoints Average Message Overhead
Cao-Singhal [4]
CaoSinghal [5]
2Tst
0h
Nmin
Nmin+ Nmut
Koo-Toeg Algorithm [11] 1
*Tch
Nmin
Elnozahy et al [8]
0h N
Proposed Algorithm 2
*Tch
Nmin
3Cbst+2Cwirele 2*Nmin*Cpp 3*Nmin*Cpp* 2*Cbst + N 2*Nmin Ndep *Cpp +Cbst *Cpp ss+2Nmss*Cst + Cbst+ +3Nmh* Cwl Nucr*Cpp
initiation of the p rocess with lo wer p rocess-ID will prevail. Table 1: A Comparison of System Performance
9. Conclusion We have proposed a m inimum process coordi nated checkpointing algorithm for m obile di stributed sy stem, where no useless checkpoi nts are t aken and an effort i s made to minimize the blocking of processes. The num ber of processes that take checkpoi nts i s m inimized t o avoi d awakening of M Hs i n doze m ode of operat ion and thrashing of MHs with checkpoi nting act ivity. Furt her, i t saves lim ited b attery life o f MHs and low bandwidth of wireless channels. W e have used t he concept of del aying selective m essages at the receiver end only during the checkpointing peri od. B y usi ng t his t echnique, only selective processes are bl ocked for a short durat ion and processes are allowed to do their normal computations and send m essages i n t he bl ocking peri od. We captured the transitive dependencies during the normal execution. The Z-dependencies are well taken care of in this protocol. We also avoi ded col lecting dependency vectors of all processes to compute t he m inimum set . Thus, t he proposed protocol is si multaneously abl e t o reduce t he useless checkpoi nts t o zero and t ries to optimize the blocking of processes at very l ess cost of maintaining exact dependencies among processes and pi ggybacking checkpoint sequence num bers and dependency vect ors onto normal computation messages.
10. References [1] Acharya A. and Badrinath B. R., “Checkpointing Distributed Applications on Mobile Com puters,” P roceedings of the 3 rd International Conference on Paralle l and Distributed Information Systems, pp. 73-80, September 1994. [2] Baldoni R., Hélary J-M., Mostefaoui A. and Ray nal M., “A Communication-Induced Checkpointing Protocol that Ensures Rollback-Dependency Trackability ,” P roceedings of the
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 www.IJCSI.org International Symposium on Fau lt-Tolerant-Computing Systems, pp. 68-77, June 1997. [3] Cao G. and Singhal M., “On coordinated checkpointing in Distributed Systems”, IEEE Transactions on P arallel and Distributed Systems, vol. 9, no.12, pp. 1213-1225, Dec 1998. [4] Cao G. and Singhal M., “On the Im possibility of Min-process Non-blocking Checkpointing and an Efficient Checkpointing Algorithm for M obile Com puting S ystems,” Proceedings of International Conference on Pa rallel Processing, pp. 37-44, August 1998. [5] Cao G. and Singhal M., “Mutable Checkpoints: A New Checkpointing Approach for Mobile Com puting systems,” IEEE Transaction On Parallel and Distri buted Systems, vol. 12, no. 2, pp. 157-172, February 2001. [6] Chandy K. M. and Lamport L., “Distributed Snapshots: Determining Global State of Distributed Sy stems,” ACM Transaction on Computing Sy stems, vol. 3, No. 1, pp. 63-75, February 1985. [7] Elnozahy E.N., Alvisi L., Wang Y.M. and Johnson D.B., “A Survey of Rollback-Recovery Protocols in Message-Passing Systems,” ACM Computing Surveys, vol. 34, no. 3, pp. 375-408, 2002. [8] Elnozahy E.N., Johnson D.B. and Zwaenepoel W., “The Performance of Consistent Checkpointing,” Proceedings of the 11th Sy mposium on Reliable Distri buted Sy stems, pp. 39-47, October 1992. [9] Hé lary J. M. , Moste faoui A. a nd Raynal M., “Communication-Induced Determination of Consistent Snapshots,” Proceedings of the 28 th International Sy mposium on Fault-Tolerant Computing, pp. 208-217, June 1998. [10] Higaki H. and Takizawa M ., “Checkpoint-recovery Protocol for Reliable Mobile Sy stems,” Trans. of Information processing Japan, vol. 40, no.1, pp. 236-244, Jan. 1999. [11] Koo R. and Toueg S., “Checkpointing and Roll-Back Recovery for Distributed S ystems,” IEEE Trans. on Software Engineering, vol. 13, no. 1, pp. 23-31, January 1987. [12] Neves N. and Fuchs W. K., “Adaptive Recovery for Mobile Environments,” Communications of the ACM, vol. 40, no. 1, pp. 68-74, January 1997. [13] Parveen Kumar, Lalit Kumar, R K Chauhan, V K Gupta “A Non-Intrusive Minimum Process Sy nchronous Checkpointing Protocol for M obile Dis tributed S ystems” P roceedings of IEEE ICPWC-2005, pp 491-95, January 2005. [14] Pradhan D.K., Krishana P.P. and Vaidya N.H., “Recovery in Mobile Wireless Environment: De sign and Trade-off Analysis,” Proceedings 26 th International Sy mposium on Fault-Tolerant Computing, pp. 16-25, 1996. [15] Prakash R. and Singhal M., “Low-Cost Checkpointing and Failure Recovery in Mobile Com puting S ystems,” IEEE Transaction On Parallel and Distri buted Systems, vol. 7, no. 10, pp. 1035-1048, October1996. [16] Ssu K.F., Yao B., Fuchs W. K. and Neves N. F., “Adaptive Checkpointing with Storage Management for Mobile Environments,” IEEE Transactions on Reliability , vol. 48, no. 4, pp. 315-324, December 1999. [17] J.L. Kim, T. Park, “An e fficient Protocol for checkpointing Recovery in Distributed S ystems,” IEEE Trans. P arallel and Distributed Systems, pp. 955-960, Aug. 1993.
32
[18] L. Kumar, M. Misra, R. C. Joshi, “Checkpointing in Distributed Computing Systems” Book Chapter “Concurrency in Dependable Computing”, pp. 273-92, 2002. [19] L. Kuma r, M. Misra , R. C. Joshi, “Low overhead optimal checkpointing for m obile distributed systems” Proceedings. 19th IEEE International Conference on Data Engineering, pp 686 – 88, 2003. [20] Ni, W., S. Vrbsky and S. Ray, “ Pitfalls in Distributed Nonblocking Checkpointing”, Journal of Interconnection Networks, Vol. 1 No. 5, pp. 47-78, March 2004. [21] L. Lamport, “Time, clocks and ordering of events in a distributed sy stem” Comm. AC M, vol.21, no.7, pp. 558-565, July 1978. [22] Silva, L.M. and J.G. S ilva, “Global checkpointing for distributed programs”, Proc. 11 th sy mp. Reliable Distributed Systems, pp. 155-62, Oct. 1992. [23] Parveen Kumar, Lalit Kumar, R K Chauhan, “ A Nonintrusive Hybrid Sy nchronous Checkpointing Protocol for Mobile S ystems”, IETE Journal of Research, Vol. 52 No. 2&3, 2006. [24] Parveen Kumar, “A Low-Cost Hy brid Coordinated Checkpointing Protocol for mob ile distributed sy stems”, To appear in Mobile Information Systems. [25] Lalit Kumar Awasthi, P.Kum ar, “A Sy nchronous Checkpointing Protocol for M obile Distributed Sy stems: Probabilistic Approach” International Journal of Inform ation and Computer Security, Vol.1, No.3 pp 298-314.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
REPUTATION BASED TWO WAY TRUST MODEL FOR RELIABLE TRANSACTIONS IN GRID COMPUTING Srivaramangai P1, Renagaramanujam Srinivasan2 1
MCA Department ,BS Abdur Rahman University Chennai , Tamilnadu, India
2
CSE Department, BS Abdur Rahman University Chennai , Tamilnadu, India
Abstract:Grids computing is a coordinated resource sharing and problem solving in any dynamic environment. Computing resources are highly heterogeneous. Grid computing and its related technologies will only be adopted by users, if they are confident that their data and privacy are secured and the system must be as scalable, robust and reliable as of their own in their places. Reliability is the probability that a process will successfully perform it’s prescribed task with out failure at a given point of time. So allowing reliable transactions plays a vital role in grid computing. Reliability of any transaction increases when it passes through two way test criteria. That is the consumer is satisfied with the ability of the provider and the provider in turn is reasonably happy with and willing to provide service to the user. To achieve this kind of reliable transactions mutual trust must be established between the initiator and the provider. Trust is measured by using reputation and reputation is the collective opinion of others. This paper provides a model which will allow only reliable transactions in grid by using trust as a measure for both provider and consumer. This enhances the security in grid.
1. Introduction: Now days the processing speed of any computer is enormously increased. Even then it is not enough to satisfy the needs of life science and physical sciences. Grid computing provides huge processing power in a distributed environment with variety of resources .The resources in Grid are shared in a flexible, coordinated and secured manner. Most of the Grid applications involve very large data bases with highly secured data. The services in grid may range from simple printing job to complex computing job. Security is one of the great issues in
grid computing. The success of grid applications depends on effective usage of resources in a way that is expected. Security mechanism in any system should prevent unauthorized entry in to the system. But in Grid environment the security should be much more than this. In grid applications the users should have reliable transactions. The reliability of any transaction is the probability of successful running or completion of a given task. So there is a need of trust system which ensures a level of robustness against malicious nodes. Trust must be established from both the sides. In this paper we analyze the existing models and identify the needs of two way reputation. We argue that every transaction must pass through two way test criteria before it gets approved. First we made analysis of existing models. Second we propose a model which improves security and allow only reliable transactions. Trust relationship is one of the prevailing security issues in Grid computing. [14] The rest of the paper is organized as follows. Section 2 discusses the related works; Section 3 proposes a new model with improved reliability; Section 4 discusses computation of trust; section 5 presents the design; and section 6 presents simulation of proposed model. In section 7 we conclude.
2. Related work: A reputation-based framework is presented by Li xiong and liu [5]. They claim that feed back values only are not enough for the calculation of trust and reputation. Y. Wang and J. Vassileva [6] propose a
33
34
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
reputation model based on Bayesian network. According to their model the peers needs are different in different situations. Selcuk et al. suggests in [7] a reputation based trust management system in which the reliability is calculated based on previous transactions. Ayman Tajeddine et al. in [8] propose a very impressive reputation based trust model. In this approach the initiator host calculates reputation value of target host based on its previous experiences and gathered feedbacks from other hosts. F.Azzedin,M.Maheswaran [9] discuss about managing trust in grid by proposing a behavior trust management model. Trust levels are graded from a to f. Both direct and indirect trust are considered. Gui Xiaolin, Xie Bing [10] propose a trust model based on behavior tracks. Attenuation function is corporated for decaying factor. Baolin Ma et al in [11] present a reputation based trusted model .Their model considers both direct feed back and feed back from other entities Direct trust is given with more weightage than the indirect score.Beulah kurian, Gregor von laszewki [12] provide a way for efficient resource selection by considering Eigen trust algorithm. Their approach is similar to Azzedin approach [9] except for a new parameter context. In our previous publication [13] the trust system is made more robust by eliminating the unreliable feedbacks by using rank correlation method.
3. Proposed model:The proposed work is an enhancement of the existing model [13] that uses rank correlation method for removing biased feed backs. It uses both direct trust and indirect trust. Direct trust is given more weightage Direct trust is calculated from the transactions which are done directly by the initiator and is given higher weightage. Indirect trust is measured by getting feed backs from entities in the same domain and also from other domains. This model calculates the credibility of the recommenders’ feedback by considering different parameters such as similarity, activity and specificity. In this model both the provider as well as the imitator collect the feed backs and the trust is calculated from both the ends. We calculate the reputation of both client as well as provider. The reputation of the client is evaluated by the provider whereas the reputation of the provider is evaluated by the client. Since the relationship between the client and provider is asymmetric and the corresponding trust reputation values will be calculated based on different parameters, the threshold values for the two will be different.
In the Earlier model the trust is calculated by the user and the decision is made based upon that trust. In the proposed model trust is measured from the provider side also and the transaction is allowed only if both the trust values are greater than a predefined threshold value. Since the reputation repository is decentralized we cannot completely depend on recommenders feed back. So here the main assumption is that there can be a few malicious entities that can give wrong feedbacks about other entities. Even if single entity is giving a wrong feedback, it is sufficient to alter the decision from one state (grant) to another (not grant) which is true for both the user and the provider. In the real world we expect a set of malicious entities trying to disrupt the smooth functioning of the grid system by false reporting giving false feedbacks. Also the entities may not be malicious but their method of evaluation may be totally different from one’s own. A, the initiator entity can evaluate the trustworthiness of provider I, based on views of colleagues, whose evaluation schemes are similar to his. The correlation can be obtained by any of the standard methods available such as Pearson Product Moment Correlation, Spearman rank Order Correlation (rho) or the Kendall rank order Correlation (tau) and we have chosen Spearman’s Rank Coefficient. In the similar way the provider also has his own right to decide whether to allow the consumer to utilize his resource or not. So he also fixes a threshold value and if the calculated trust is greater than that values he allows the transaction else denies. Thus even if the initiator is ready to take up a particular resource from the provider the provider also should be in the position to share his resource. This is a two way test criteria which assures more reliable transaction which in turn improves security in grid environment. In this model every entity can be either a provider or a user for any particular transaction.
4. Computation of Trust: The computation of trust is depicted in this section .Let us assume A is the provider and I is the user. The model decides the transaction as follows; First I as the initiator has to decide whether to accept the resource from A or not. The total trust (trust 1) is calculated by the expression; Total trust = u*direct trust + v * indirect 1 + w * indirect 2 (1) Where u+v+w=1 and u>v>w. αi rep y/xi ij indirect 1=
αi
(2)
35
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
ik
indirect 2 =
βj rep y/xi jk βj jk
the provider entities are considered not the users, trust2. If Trust 1 > γ (minimum threshold value) and trust 2 > δ then the transaction is allowed else the transaction is denied.
(3)
α, β are credibility factors . Direct trust is the value which is perceived by the initiator entity by his own experience. Indirect 1 & indirect 2 both are calculated by taking the recommenders’ feed backs. In indirect 1 the feedbacks are collected from its neighbours. That is the entities from the same domain. In indirect 2 the feedbacks are collected from the entities in foreign domains. In this model the similarity between the requester and each recommender is estimated by rank correlation method. (Spearman rank Order Correlation (rho)). If the correlation is greater than zero then the entity’s feedback is taken, thus avoiding biased feed backs. Credibility of each recommender is measured by using similarity, activity and specificity. [14] The following data gives one sample output which explains the allocation procedure. Over all fifteen entities have been taken. Two domains are considered. The first model takes all the feed backs. The second and third model takes only the reliable feedbacks. G and O are both reputed. So the transaction is rightly granted for them by the proposed model. Entities considered: [A, B, C, D, E, F, G, H, I, J, K, L, M, N, O] Domain1:[A, B, C, D, E, F, G] Domain2:[H, I, J, K, L, M, N, O] Malicious = [C,F,K,N]
5. Design of the proposed system:
Fig1 explains the overall architecture. The user uses the reputation model and decides whether to choose a particular provider or not. In turn providers also use the model with its own recommendations and make decisions. Provide feedback about provider
user
Two ways Reputation Model
Results
Provide r
Results
Fig 1 Context diagram for the proposed model. Fig 2 explains about the calculation of user side trust. The user get the recommendation from other entities, calculate the rank correlation and decides whether to accept the recommendations. If the rank correlation is positive he accepts. Otherwise he goes to the next feedback. Then he calculates the total trust. If the total trust is greater than the minimum threshold he chooses that provider. Fig 3 depicts the procedure for provider side trust.
EXISTING MODEL [13] Initiator G Provider: O Recommenders:[A, B, C, D, E, F, H, I, J, K, L, M, N] Total Trust Score: 2.92 Resource is not allocated PROPOSED MODEL Initiator: G Provider: O Recommenders:[A, B, D, E, H, I, J, L] (only reliable nodes ) Total Trust Score: 3.13 Resource is Allocated In the same way the total trust about the user is measured by the provider. But this time only
Provide feedbac k about User
Fig 2 Procedure for user side trust
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
Fig 3 Procedure for provider side trust
Fig 4 – Flow chart for the overall process
36
37
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
The above flowchart explains the system flow as a whole. First the initiator entity I choose a provider P. It gets the recommendations from all of its neighbours. Credibility of feedbacks are calculated and based upon the credibility value the initiator considers the feed backs which is reliable. Then total trust and indirect trust are calculated using the equations 1, 2 & 3 . If the trust value is greater than the chosen threshold then the user requests the resource from the provider. Provider goes through all the steps and if the calculated trust value satisfies his threshold he agrees to offer his resource else he sets the status as busy. If both the provider and consumer agree upon their trust value the transaction is executed.
6. Experiments and results: Simulation studies have best conducted using three models as follows. Model 1: Existing model as proposed by [8] Model2: Model proposed by us earlier [13], which depends on feedbacks from users only. Any initiator obtains feedbacks from other users, eliminates biased/malicious feedbacks and makes a decision based upon trust evaluated from remaining feed backs. Model3: Present proposal, which is an improvement over model 2. Here providers also get feedback from other providers about the user /initiator who is making request for services. Identification of malicious feedbacks is done using ranking coefficients. Thus the two ways model eliminates biased feedbacks from both users and providers. Three sets of simulation studies have been done each with 100 runs. The results are presented in Table 1 and 2 and Fig. 5. Table 1 provides the cumulative summary of all the three simulations while Table 2 presents detailed analysis of disagreement cases corresponding to a particular study – study 2. From a perusal of Table 1 , We find that there are altogether eight combinations of results . YYY signifies that all the three models agree to grant a request by the initiator while NNN denotes that all the models reject the request. We find that out of a total of 300 instances there is complete agreement among all the three models in 258 (138+120) cases constituting 86% of runs. The disagreement in 42 cases can be spilt into six categories. Let us focus our attention to each one of these categories. Column 4 corresponds to YNN situation where Model 1 grants request while Models 2 and 3 refuse. We have checked each one and have found Models 2 and 3 have correctly eliminated malicious / biased provider,
Columns 5 & 6 (NYN, YYN) correspond to a situation where providers assessments about the initiators/users based on feedbacks given by the other providers, fail to meet the minimum threshold limits. That means the providers are not satisfied with the trustworthiness of the users and fail to do business with them. In all there are 14 instances corresponding to these combinations, constitute 4.7 % of the total instances. Column 7 (NYY) pertains to the situation where Model 1 rejects while Models 2 and 3 grant the request. The numbers of instances are 25 constituting 8.3% of the population. The first model has taken the biased feedbacks also in to calculation of trust and wrongly rejects the request. Models 2 & 3 have arrived at the correct decision of granting the request after successfully eliminated the malicious nodes. Columns 8 and 9, corresponding to YNY and NNY contain zero entries since these correspond to the cases where the second model rejects, since threshold limits are not satisfied regarding a provider, so also does Model 3 which is a superset of Model 2. Table 1: Cumulative results
S. N O 1 2 3 4 5 6 7 8 9 10 11 12 13 14
User Pr C K C C K C N B E G K B G L
ovider J F D J B J K I O H C I H E
Model1
Model2
Model3
YES
YES
NO
NO NO
YES YES
NO YES
Table 2 – Disagreement cases Sim ulat ion 1. 2. 3. Tot Per
YY Y
NN N
41 46 51 138 46
44 40 36 120 40
Y N N 2 0 1 3 1
NY N 1 1 3 5 1.6
Y Y N 1 6 2 9 3
NY Y 11 7 7 25 8.4
Y N Y 0 0 0 0 0
N N Y 0 0 0 0 0
Tota l Runs 100 100 100 300
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
In the first set of results that is no 1 to 6 the first two models accept the transaction. But the proposed model denies the transaction. In all the six cases the user is malicious. The proposed model considers the reputation from both the sides. It denies the transaction by applying the two way test criteria. The malicious nodes are prevented from accessing the resources there by the reliability of transaction are improved in the proposed model. In transaction 7 the user is malicious. In Model 3 the provider knows the malicious behavior of users. So it correctly denies the resource. In transaction 11 both the user and the provider are malicious. So they agree upon each other and the transaction begins. Other transactions 8 to14 both have good reputations and hence the transaction is executed. But Model 1 wrongly denies the transaction for the right nodes. Out of all 14 disagreement cases our proposed model makes right decision with the malicious nodes.
Fig 5.c: Allocations for three simulations Another analysis of the results brings out interesting information. From Table 1 we find model 1 grants 150 (138+3+9) of the 300 requests by the initiator, while model 2 grants 177 (138+5+9+25) of the total requests. For model 3 we find that in all 163 (138+25) requests are sanctioned. Initially when we design the two way trust model we felt through put as measured by the number of requests granted as compared to the requests made will come down because of the two level filtering features – one at the provider and the other at the user level. This has not happened because model 1 the existing model is affected by the biased feedbacks, and has arrived at wrong decision for 25 cases. Fig 6 brings out this aspect.
Fig 5.a: Allocations for three simulations
Fig 6 Comparison of Allocation.
Fig 5.b: Allocations for three simulations
Another aspect to be considered is the accuracy of the models – What percentage of each model decision is correct. We find that Models 1,2 and 3 have accuracy values of (258/300) 86%, (286/300) 95.33% and (300/300) 100% respectively . Thus Model 3 arrives at the correct decision in all the cases
38
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
7. Conclusion: The paper has presented a new trust model which is comprehensive in the sense it takes cognizance of both provider and user sensibilities. Further by eliminating biased feedbacks from both user and provider groups the resultant transactions become more reliable and secure .Simulation study establish the superiority of the proposed two way trust model over the existing models. Acknowledgements The authors would like to express their thanks to Dr. P. Kanniappan, Vice chancellor, Dr.V.M.Periasamy , Registrar, Dr.K.M.Mehata the Dean and Dr.P.Sheik abdul Khader , HOD/Dept of Computer Applications of B.S.Abdur Rahman University for the environment provided. References: [1] A.Arenas “State of art survey on trust and security in grid computing system “ march 2006. [2] Gheorghe Cosmin Silaghi, Alvaro E. Arenas, Luis Moura Silva ,” Reputation-based trust management systems and their applicability to grids “CoreGRID Technical Report Number TR0064 February 23, 2007 [3] Marcim Adamski, Alvaro Arenas, Angelos Bilas “ “Trust and Security in Grids: A State of the Art” CoreGRID White Paper Number WHP-0001 May 26, 2008 [4] A. Abdul-Rahman and S. Hailes. “Supporting trust in virtual communities”. In HICSS ’00: Proceedings of the 33rd Hawaii International Conference on System Sciences-Volume 6, page 6007, Washington, DC, USA, 2000. IEEE Computer Society. [5] L. Xiong, and L. Liu, “PeerTrust: Supporting Reputation-Based Trust for Peer-to-Peer Electronic
Communities,” IEEE Transactions on Knowledge and Data Engineering, Vol. 16, No. 7, July 2004. [6] Y. Wang and J. Vassileva, “Trust and Reputation Model in Peer-to-Peer Networks,” Proceedings of the Third International Conference on Peer-to-Peer Computing (P2P’03), 2003. [7] A. Selcuk, E. Uzun, and M. Pariente, “A Reputation-Based Trust Management System for P2P Networks,” IEEE International Symposium on Cluster Computing and the Grid 2004. [8] Ayman Tajeddine Ayman Kayssi Ali Chehab Hassan Artail “A Comprehensive ReputationBased Trust Model for Distributed Systems “ IEEE 2005. [9] F.Azzedin,M.Maheswaran “Evolving and managing trust in grid computing system” IEEE CCECE,2002. [10] Gui Xiaolin, Xie Bing “Study on behavior based trust model in grid security system” “Proceedings of the 2004 IEEE International Conference on Services Computing (SCC’04). [11] Baolin Ma, Jizhou Sun, Ce Yu ” Reputationbased Trust Model in Grid Security System “ , Aug. 2006, Volume 3, No.8 (Serial No.21) Journal of Communication and Computer, ISSN1548-7709, USA. [12] Beulah kurian, Gregor von laszewki “Reputation based grid resource selection” [13] P.Srivaramangai , .R.Srinivasan “ Reputation based trust model with elimination of unreliable feedbacks “ International journal of information technology and knowledge management Vol2 , 2009. No 2,pp.455-459 [14] P.Srivaramangai , R.Srinivasan “ An overview of Trust model for Grid Security” National conference ,March 2007.
39
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
40
Soft-Checkpointing Based Coordinated Checkpointing Protocol for Mobile Distributed Systems Parveen Kumar1, Rachit Garg2 1
Meerut Institute of Engineering & Technology, Department of Computer Science & Engineering, Meerut (INDIA)- 250005 2 Singhania University, Department of Computer Science & Engineering, Pacheri Bari (Rajasthan), India
Abstract: Minimum-process coordinated checkpointing is a
as a set of local states, one from each process. The state of channels
suitable approach to introduce fault tolerance in mobile distributed
corresponding to a global state is the set of messages sent but not
systems transparently. It may require blocking of processes, extra
yet received. A global state is said to be “consistent” if it contains
synchronization messages or taking some useless checkpoints. All-
no orphan message; i.e., a message whose receive event is recorded,
process checkpointing may lead to exceedingly high checkpointing
but its send event is lost [5]. To recover from a failure, the system
overhead. To optimize both matrices, the checkpointing overhead
restarts its execution from the previous consistent global state saved
and the loss of computation on recovery, we propose a hybrid
on the stable storage during fault-free execution. This saves all the
checkpointing algorithm, wherein an all-process coordinated
computation done up to the last checkpointed state and only the
checkpoint is taken after the execution of minimum-process
computation done thereafter needs to be redone. In
coordinated checkpointing algorithm for a fixed number of times. In
coordinated
or
synchronous
checkpointing,
the minimum-process coordinated checkpointing algorithm; an
processes take checkpoints in such a manner that the resulting
effort has been made to optimize the number of useless checkpoints
global state is consistent. Mostly it follows the two-phase commit
and blocking of processes using probabilistic approach and by
structure [2], [5], [6], [7], [10], [15]. In the first phase, processes
computing an interacting set of processes at beginning. We try to
take tentative checkpoints, and in the second phase, these are made
reduce the loss of checkpointing effort when any process fails to
permanent. The main advantage is that only one permanent
take its checkpoint in coordination with others. We reduce the size
checkpoint and at most one tentative checkpoint is required to be
of checkpoint sequence number piggybacked on each computation
stored. In the case of a fault, processes rollback to the last
message.
checkpointed state [6]. The Chandy-Lamport [5] algorithm is the earliest
1. Introduction
non-blocking
all-process
coordinated
checkpointing
algorithm.
Mobile Hosts (MHs) are increasingly becoming common in
The existence of mobile nodes in a distributed system
distributed systems due to their availability, cost, and mobile
introduces new issues that need proper handling while designing a
connectivity. An MH is a computer that may retain its connectivity
checkpointing algorithm for such systems [1], [4], [14], [16]. These
with the rest of the distributed system through a wireless network
issues are mobility, disconnections, finite power source, vulnerable
while on move. An MH communicates with the other nodes of the
to physical damage, lack of stable storage etc. Prakash and Singhal
distributed system via a special node called mobile support station
[14] proposed a nonblocking minimum-process coordinated
(MSS).
A “cell” is a geographical area around an MSS in which it
checkpointing protocol for mobile distributed systems. They
can support an MH. An MSS has both wired and wireless links and
proposed that a good checkpointing protocol for mobile distributed
it acts as an interface between the static network and a part of the
systems should have low overheads on MHs and wireless channels;
mobile network. Static nodes are connected by a high speed wired
and it should avoid awakening of an MH in doze mode operation.
network [1].
The disconnection of an MH should not lead to infinite wait state. A checkpoint is a local state of a process saved on the
The algorithm should be non-intrusive and it should force minimum
stable storage. In a distributed system, since the processes in the
number of processes to take their local checkpoints. In minimum-
system do not share memory, a global state of the system is defined
process coordinated checkpointing algorithms, some blocking of the
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
41
processes takes place [3], [10], [11], or some useless checkpoints
separated sequential processes denoted by P0, P1, ..., Pn-1, running
are taken [4], [15].
on fail-stop MH’s or on MSS’s. Each MH or MSS has one process
In
minimum-process
coordinated
checkpointing
running on it. The processes do no share common memory or
algorithms, a process Pi takes its checkpoint only if it a member of
common clock. Message passing is the only way for processes to
the minimum set (a subset of interacting process). A process Pi is in
communicate with each other. Each process progresses at its own
the minimum set only if the checkpoint initiator process is
speed and messages are exchanged through reliable channels,
transitively dependent upon it. Pj is directly dependent upon Pk only
whose transmission delays are finite but arbitrary. We assume the
if there exists m such that Pj receives m from Pk in the current
processes to be non-deterministic.
checkpointing interval [CI] and Pk has not taken its permanent checkpoint after sending m. The ith CI of a process denotes all the computation performed between its ith and (i+1)th checkpoint, including the ith checkpoint but not the (i+1)th checkpoint. Cao and Singhal [4] achieved non-intrusiveness in the minimum-process algorithm mutable
by introducing the concept of
checkpoints. Kumar and Kumar [21] proposed a
minimum-process coordinated checkpointing algorithm for mobile distributed systems, where the number of useless checkpoints and the blocking of processes are reduced using a probabilistic approach. Singh and Cabillic [20] proposed a minimum-process non-intrusive coordinated checkpointing protocol for deterministic mobile systems, where anti-messages of selective messages are logged during checkpointing. Higaki and Takizawa [8], and Kumar et al [17] proposed hybrid
checkpointing protocols where MHs
checkpoint independently and MSSs checkpoint synchronously. Neves et al. [13] gave a time based loosely synchronized coordinated checkpointing protocol that removes the overhead of synchronization and piggybacks integer csn (checkpoint sequence number). Pradhan et al [19] had shown that asynchronous checkpointing with message logging is quite effective for checkpointing mobile systems. In the present study, we present a hybrid scheme, where an all process checkpoint is enforced after executing minimum-process algorithm for a fixed number of times as in [22]. In the first phase, the MHs in the minimum set are required to take soft checkpoint only. Soft Checkpoint is stored on the disk of the MH and is similar to mutable checkpoint [4]. In the minimumprocess algorithm, a process takes its forced checkpoint only if it is having a good probability of getting the checkpoint request; otherwise, it buffers the received messages as in [21].
2. The Proposed Checkpointing Algorithm 2.1 System Model
2.2 Basic Idea Similar to [3], [21], [22], initiator process collects the dependency vectors of all processes and computes the tentative minimum set. Suppose, during the execution of the checkpointing algorithm, Pi takes its checkpoint and sends m to Pj. Pj receives m such that it has not taken its checkpoint for the current initiation and it does not know whether it will get the checkpoint request or not. If Pj takes its checkpoint after processing m, m will become orphan. In order to avoid such orphan messages, we use
the following technique as
mentioned in [21]. If Pj has sent at least one message to a process, say Pk, and Pk is in the tentative minimum set, there is a good probability that Pj will get the checkpoint request. Therefore, Pj takes its mutable checkpoint before processing m [4]. In this case, most probably, Pj will get the checkpoint request and its mutable checkpoint will be converted into permanent one. Alternatively, this message is buffered Pj. Pj will process m only after taking its tentative checkpoint or after getting commit as in [22]. In minimum-process checkpointing, some processes may not be included in the minimum set for several checkpoint initiations due to typical dependency pattern; and they
may starve for
checkpointing. In the case of a recovery after a fault, the loss of computation at such processes may be unreasonably high [22]. In Mobile Systems, the checkpointing overhead is quite high in allprocess checkpointing [14]. Thus, to balance the checkpointing overhead and the loss of computation on recovery, we design a hybrid checkpointing algorithm for mobile distributed systems, where an all-process checkpoint is taken after certain number of minimum-process
checkpoints.
We
enforce
the
all-process
checkpointing protocol after executing the minimum-process algorithm for fifteen number of times. In coordinated checkpointing, if a single process fails to take its checkpoint; all the checkpointing effort goes waste, because, each process has to abort its tentative checkpoint. In order
The system model is similar to [3], [4]. A mobile computing system
to take the tentative checkpoint, an MH needs to transfer large
consists of a large number of MH’s and relatively fewer MSS’s. The
checkpoint data to its local MSS over wireless channels. Hence, the
distributed computation we consider consists of n spatially
loss of checkpointing effort may be exceedingly high. Therefore, we
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
42
propose that in the first phase, all concerned MHs will take soft
to the minimum set; computation of minimum set
checkpoint only. Soft checkpoint is similar to mutable checkpoint
on the bases of dependency vectors of all processes
[4], which is stored on the memory of MH only. In this case, if some
is given in [21].
process fails to take checkpoint in the first phase, then MHs need to
R1[]:
abort their soft checkpoints only. The effort of taking a soft checkpoint is negligible as compared to the tentative one. When the
taken its soft checkpoint in the first phase;
R2[]:
initiator comes to know that all relevant processes have taken their soft checkpoints, it asks all relevant processes to come into the
a bit vector of length n; R[i] =1 implies Pi has a bit vector of length n; R2[i] =1 implies Pi has taken its tentative checkpoint in the second phase;
Timer1:
a flag; initialized to ‘0’ when the timer is set; set
second phase, in which, a process converts its soft checkpoint into
to ‘1’ when
tentative one. Finally, the initiator issues the commit request.
collecting coordinated checkpoint expires;
2.3 Data Structures Here, we describe the data structures used in the proposed checkpointing protocol. A process that initiates checkpointing is called initiator process and its local MSS is called initiator MSS. Data structures are initialized on completion of a checkpointing
iii) MSS (say MSSp) maintains the following data structures:
mss_localp[]:
also includes the processes running on MH’s, which have been
mss_loc_tentp[]:
tentative checkpoint at MSSp;
mss_loc_softp[]:
Each process Pi maintains the following data structures, which
d_vecti[]:
soft checkpoint in the first phase and Pi is local to MSSp;
mss_tent_reqp[]:
a flag which indicates that Pi is in blocking state;
pr_c_statei:
checkpoint request has been sent to process Pi and Pi is local to MSSp;
mss_soft_reqp[]:
request has been sent to process Pi in the first phase and Pi is local to
during checkpointing; a flag; set to ‘1’ on mutable checkpoint; reset on commit/abort or on tentative checkpoint;
pr_sendvi[]:
CI;
pr_sendi:
MSSp;
mss_fail_bit:
a flag indicating that Pi has sent at least one
Pin:
initiator process identification;
g_chkpt:
a flag; set to ‘1’ on the receipt of ddv[] request; It controls multiple checkpoint initiations;
four bits checkpoint sequence no; initially, for a process pr_csn and pr_next_csn are [0000] and
rec_mset
commit/abort;
pr_csn=pr_next_csn;
pr_next_csn=modulo 16 (++pr_next_csn) ;
a flag; set to 1 on the receipt of mset[] from the initiator MSS; set to ‘0’ on
[0001] respectively; pr_csn is incremented as follows:
to ‘1’ when some relevant
checkpoint;
message since last checkpoint;
pr_csn
a flag; set
process in its cell fails to take its
a bit vector of size n; pr_sendvi[j]=1 implies Pi has sent at least one message to Pj in the current
a bit vector of length n; mss_soft_reqp[i]=1 implies soft checkpoint
a flag; set to ‘1’ on soft or mutable checkpoint or on the receipt of a message of higher csn
mutablei:
a bit vector of length n; mss_tent_reqp[i]=1 implies tentative
a bit vector of size n; ; d_vecti[j] =1 implies Pi is directly dependent upon Pj for the current CI;
pr_blocki:
a bit vector of length n; mss_loc_softp[i]=1 implies Pi has taken its
is still with this MSS. are preferably stored on local MSS:
a bit vector of length n; mss_loc_tentp[i]=1 implies Pi has taken its
disconnected from the MSS but their checkpoint related information i)
a bit vector of length n; mss_localp [i]=1 implies Pi is running in the cell of MSSp;
process if not mentioned explicitly. A process is in the cell of an MSS if it is running on the MSS or on an MH supported by it. It
maximum allowable time for
new_set[]
a bit vector of length n; it contains all new processes found for the minimum set at the MSS; on each checkpoint request: if
ii) Initiator MSS (any MSS can be initiator MSS) maintains the
(tnew_set≠) new_set=new_settnew_set;
following Data structures:
mset[]:
a bit vector of size n; mset[k]=1 implies Pk belongs
Tnew_set[]
a bit vector of length n; it contains the new
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
43
processes found for the minimum set while
the time of receiving m11, P3 has received the mset[] and it P3 is not
executing a particular checkpoint request.
the member of the tentative minimum set (mset[]). P3 finds that it
When a process, say Pi, takes its soft
has sent m3 to P2 and P2 is a member of tentative minimum set
checkpoint, it may find some process Pj
(mset[]). Therefore, P3 concludes that most probably, it will get the
such that Pi is dependent upon Pj and Pj is
checkpoint request in the current initiation; therefore, it takes its
not in the tentative minimum set known to
mutable checkpoint before processing m11. When P2 takes its soft
the local MSS;
checkpoint, it finds that it is dependent upon P3, due to m3, and P3 is not in the tentative minimum set [mset[]]; therefore, P2 sends checkpoint request to P3. On receiving the checkpoint request, P3
2.4 An Example checkpointing
converts its mutable checkpoint into soft one. It should be noted that
algorithm with the help of an example. In Figure 1, at time t1, P1
the soft checkpoint and mutable checkpoint are similar. Mutable
initiates checkpointing process and sends request to all processes for
checkpoint is a forced checkpoint and soft checkpoint is a regular
their dependency vectors. At time t2, P1 receives the dependency
checkpoint taken due to checkpoint request. In order to convert the
vectors from all processes and computes the tentative minimum
mutable checkpoint into soft checkpoint, we only need to change
(mset[]) set as in [21], which in case of Figure 1 is {P0, P1, P2}]. P1
the data structure (mss_local_soft[3]=1).
We
explain
the
proposed
minimum-process
sends this tentative minimum set to all processes and takes its own soft checkpoint. A process takes its soft checkpoint if it is a member
After taking its checkpoint, P2 sends m13 to P4. P4 finds that it has
of the tentative minimum set. When P0 and P2 get the mset[], they
not sent any message to a process of tentative minimum set. It takes
find themselves in the mset[]; therefore, they take their soft
the bitwise logical AND of pr_sendv4[] and mset[] and finds the
checkpoints. When P3, P4 and P5 get the mset[], they find that they
resultant
do not belong to mse[], therefore, they do not take their soft
mset[]=[111000]). P4 concludes that most probably, it will not get
checkpoints.
the checkpoint request in the current initiation; therefore, P4 does
vector
to
be
all
zeroes
(pr_sendv4[]=[000001];
not take mutable checkpoint but buffers m13. P4 processes m13 only after getting the tentative checkpoint request. P5 processes m14, because, it has not sent any message since last permanent checkpoint (pr_send5=0) . After taking its checkpoint, P1 sends m12 to P2. P2 processes m12, because, it has already taken its checkpoint in the current initiation. At time t3, P1 receives positive responses to soft checkpoint requests from all relevant processes (not shown in the Figure 1) and issues tentative checkpoint request along with the exact minimum set [P0, P1, P2, P3 ] to all processes. On receiving tentative checkpoint request, all relevant processes convert their soft checkpoints into tentative ones and inform the initiator. A process, not in the minimum set, discards its mutable checkpoint, if any, or processes the buffered messages, if any. Finally, at time t4, initiator P2 issues commit.
On receiving commit following actions are
taken. A process, in the minimum set, converts its tentative P1 sends m8 after taking its soft checkpoint and P0 receives m8 before getting the mset[].When P1 sends m8 to P0, P1 also piggybacks pr_csni and pr_c_statei along with m. When P0 receives m8 it finds that csn[i] csn[j] ; otherwise, it simply processes the message.
pr_own_csn=1). If ((m.pr_c_state) (pr_sendi==1) (rec_mset==1) ( Bitwise
3. Conclusions
logical AND of mset[] and pr_sendvi[] is all zeroes )) then ( Pi sets
We propose a hybrid checkpointing algorithm, wherein, an all-
pr_blocki flag and buffers m).
process coordinated checkpoint is taken after the execution of
Case4. Pi has entered the checkpoint state (pr_c_statei=1)
minimum-process coordinated checkpointing algorithm for a fixed
Pi processes m.
number of times. In minimum-process checkpointing, we try to
2.5.4
reduce the number of useless checkpoints and blocking of
Termination
When an MSS learns that all of its relevant processes have taken their soft/tentative checkpoints successfully or at least one of its relevant process has failed to take its soft/tentative checkpoint, it sends the response message along with new_set
to the initiator
MSS. If, after sending the response message, an MSS receives the checkpoint request along with tnew_set, and learns that there is at least one process in the tnew_set running in its cell and it has not taken the soft checkpoint, the MSS requests such process to take checkpoint. It again sends the response message to the initiator MSS. Initiator MSS commits only if every relevant process takes its tentative checkpoint. When the initiator MSS receives a response from some
processes. We have proposed a probabilistic approach to reduce the number of useless checkpoints. Thus, the proposed protocol is simultaneously able to reduce the useless checkpoints and blocking of processes at very less cost of maintaining and collecting dependencies and piggybacking checkpoint sequence numbers onto normal messages. Concurrent initiations of the proposed protocol do not cause its concurrent executions. We try to reduce the loss of checkpointing effort when any process fails to take its checkpoint in coordination with others.
References 1)
MSS, it updates its mset[] on the basis of new_set received with the response. Finally, initiator MSS sends commit/abort to all processes. On receiving commit: if a process, say Pi, belongs to the minimum
2)
set, it converts its tentative checkpoint into permanent one and discards its earlier permanent checkpoint, if any; otherwise, it processes the buffered messages, if any or discards its mutable
3)
checkpoint, if any.
2.6 All Process Checkpointing Algorithm 4) Our all process checkpointing algorithm is an updating of Elnozahy et al [7]. Initiator MSS sends soft checkpoint request to all MSSs. On receiving the soft checkpoint request, an MSS sends the request
5)
to all processes in its cell. A process takes its soft checkpoint if it has not taken the same during the current initiation. A process, after taking its tentative checkpoint or knowing its inability to take the
6)
checkpoint, informs its local MSS. When an MSS learns that all of its processes have taken their soft checkpoints, it informs the initiator MSS. When the initiator MSS receives positive response
7)
from all MSSs, it issues tentative checkpoint request to all MSSs. If any process fails to take soft checkpoint, initiator MSS issues abort request. Finally, initiator MSS issues commit request. When a process sends a computation message, it appends its pr_csn with the message. When a process, say Pi,
8)
Acharya A. and Badrinath B. R., “Checkpointing Distributed Applications on Mobile Computers,” Proceedings of the 3rd International Conference on Parallel and Distributed Information Systems, pp. 73-80, September 1994. Cao G. and Singhal M., “On coordinated checkpointing in Distributed Systems”, IEEE Transactions on Parallel and Distributed Systems, vol. 9, no.12, pp. 1213-1225, Dec 1998. Cao G. and Singhal M., “On the Impossibility of Minprocess Non-blocking Checkpointing and an Efficient Checkpointing Algorithm for Mobile Computing Systems,” Proceedings of International Conference on Parallel Processing, pp. 37-44, August 1998. Cao G. and Singhal M., “Mutable Checkpoints: A New Checkpointing Approach for Mobile Computing systems,” IEEE Transaction On Parallel and Distributed Systems, vol. 12, no. 2, pp. 157-172, February 2001. Chandy K. M. and Lamport L., “Distributed Snapshots: Determining Global State of Distributed Systems,” ACM Transaction on Computing Systems, vol. 3, No. 1, pp. 6375, February 1985. Elnozahy E.N., Alvisi L., Wang Y.M. and Johnson D.B., “A Survey of Rollback-Recovery Protocols in MessagePassing Systems,” ACM Computing Surveys, vol. 34, no. 3, pp. 375-408, 2002. Elnozahy E.N., Johnson D.B. and Zwaenepoel W., “The Performance of Consistent Checkpointing,” Proceedings of the 11th Symposium on Reliable Distributed Systems, pp. 39-47, October 1992. Higaki H. and Takizawa M., “Checkpoint-recovery Protocol for Reliable Mobile Systems,” Trans. of Information processing Japan, vol. 40, no.1, pp. 236-244, Jan. 1999.
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
9)
10)
11)
12)
13)
14)
15)
16)
17)
18)
19)
20)
21)
22)
J.L. Kim, T. Park, “An efficient Protocol for checkpointing Recovery in Distributed Systems,” IEEE Trans. Parallel and Distributed Systems, pp. 955-960, Aug. 1993. Koo R. and Toueg S., “Checkpointing and Roll-Back Recovery for Distributed Systems,” IEEE Trans. on Software Engineering, vol. 13, no. 1, pp. 23-31, January 1987. Parveen Kumar, R K Chauhan, “A Coordinated Checkpointing Protocol for Mobile Computing Systems”, International Journal of Information and Computing Science, Vol. 9, No. 1, pp. 18-27, 2006. Lalit Kumar, M. Misra, R.C. Joshi, “Low overhead optimal checkpointing for mobile distributed systems” Proceedings. 19th International Conference on IEEE Data Engineering, pp 686 – 88, 2003. Neves N. and Fuchs W. K., “Adaptive Recovery for Mobile Environments,” Communications of the ACM, vol. 40, no. 1, pp. 68-74, January 1997. Prakash R. and Singhal M., “Low-Cost Checkpointing and Failure Recovery in Mobile Computing Systems,” IEEE Transaction On Parallel and Distributed Systems, vol. 7, no. 10, pp. 1035-1048, October1996. Weigang Ni, Susan V. Vrbsky and Sibabrata Ray, “ Pitfalls in nonblocking checkpointing” World Science’s journal of Interconnected Networks. Vol. 1 No. 5, pp. 4778, March 2004. Parveen Kumar, Lalit Kumar, R K Chauhan, “A low overhead Non-intrusive Hybrid Synchronous checkpointing protocol for mobile systems”, Journal of Multidisciplinary Engineering Technologies, Vol.1, No. 1, pp 40-50, 2005. Lalit Kumar, Parveen Kumar, R K chauhan “Logging based Coordinated Checkpointing in Mobile Distributed Computing Systems”, IETE journal of research, vol. 51, no. 6, 2005. Lamports L., “Time, clocks and ordering of events in distributed systems” Comm. ACM, 21(7), 1978, pp 558565. Pradhan D.K., Krishana P.P. and Vaidya N.H., “Recovery in Mobile Wireless Environment: Design and Trade-off Analysis,” Proceedings 26th International Symposium on Fault-Tolerant Computing, pp. 16-25, 1996. Pushpendra Singh, Gilbert Cabillic, “A Checkpointing Algorithm for Mobile Computing Environment”, LNCS, No. 2775, pp 65-74, 2003. Lalit Kumar Awasthi, P.Kumar, “A Synchronous Checkpointing Protocol for Mobile Distributed Systems: Probabilistic Approach” International Journal of Information and Computer Security, Vol.1, No.3 pp 298314, 2007. Parveen Kumar, “A Low-Cost Hybrid Coordinated Checkpointing Protocol for Mobile Distributed Systems”, Mobile Information Systems pp 13-32, Vol. 4, No. 1, 2007.
46
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
47
Footprint Recognition using Modified Sequential Haar Energy Transform (MSHET) V. D. Ambeth Kumar1 1Research
M. Ramakrishnan2
scholar in sathyabamauniversity,
Chennai, Tamil Nadu- 600 119, India.
2Professor
and Head of Department Of Information Technologies,
Velammal Engineering College, Chennai, Tamil Nadu.
Abstract
Footprint identification is the measurement of footprint features for recognizing the identity of a user. Footprint is universal, easy to capture and does not change much across time. Footprint biometric system does not require specialized acquisition devices. Footprint image of a left leg is captured for hundred people in different angles. No special lighting is used in this setup. The foot image is positioned and cropped according to the key points. Sequential modified Haar transform is applied to the resized footprint image to obtain Modified Haar Energy (MHE) feature. The sequential modified Haar wavelet can map integer-valued signals onto integer-valued signals abandoning the property of perfect reconstruction. The MHE feature is compared with the feature vectors stored in database using Euclidean Distance. The accuracy of the MHE feature and Haar energy feature under different decomposition levels and combinations are compared. Above 92.375% percent accuracy can be achieved using proposed MHE feature.
Keywords: Foot print, Sequential Haar Transform network, Heel Shape, Information Security
1. Introduction A lot of automated biometrics based identification and verification systems have been developed [1]. The biometrics features derived from fingerprints [2], faces [2], irises, retinas, a speaker’s voice, and perhaps a variety of other characteristics. The systems are now used in a wide range of environments, such as law enforcement, social welfare, banking, and various security applications.
Some systems for personal identification use fundamental biometric features derived from fingerprints and irises. To acquire the feature, subjects must input their biometrics to a sensor. Signatures [3] and speakers’ voices [4] are also features useful for verification; however, obtaining these features requires the subjects’ cooperation. The main problem in automatic personal identification is how to verify the sampled feature against the registered feature with high reliability. Practical methods for automated biometrics based identification have not yet been developed for use with unconstrained subjects. Problem in automatic personal identification is how to verify the sampled feature against the registered feature with high reliability. Practical methods for automated biometrics based identification have not yet been developed for use with unconstrained subjects. A high recognition rate by verifying raw footprints directly is difficult to obtain, because people stand in various positions with different distances and angles between the two feet. To achieve robustness in matching an input pair of footprints with those of registered footprints, the input pair of footprints must be normalized in position and direction. Such normalization might remove useful information for recognition, so geometric information of the footprint prior to normalization into an evaluation function for personal recognition decision is included. In this paper, we propose a footprint-based personal recognition method and test its reliability. Footprint texture features are usually extracted using transform-based method such as Fourier Transform [5] and Discrete Cosine Transform [6]. Besides that,
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
Wavelet Transform [7] is also used to extract the texture features of the footprint. In this work, a Sequential modified Haar Wavelet is proposed to find the Modified Haar Energy (MHE) feature. Fig 1,2 shows the proposed footprint identification using Sequential modified Haar Transform. Footprint
Image Cropping
Image acquisition
RGB to Gray Scale conversion
48
features of the footprint are extracted using Sequential modified Haar Wavelet. The Modified Haar Energy (MHE) is represented using feature vector and compared using Euclidean Distance with the threshold values stored in the database. Fig 2 shows the process of footprint identification using feature extraction. The footprint image captured using digital camera is extracted for its features. The already existing features of footprint images stored in database are compared. Then, the footprint image is accepted or rejected based on the comparison.
2. Proposed Method To obtain a footprint image, from digital camera. Method for acquiring the footprints are described in the next section. This section, we describe a normalization procedure and a recognition method.
2.1 Normalization Modified Haar Energy (MHE)
Transformation Techniques
LPCC and LPC
Database storage Fig. 1 Footprint identification using Sequential Modified Haar Wavelet
2.1.1. Image Acquisition Hundred left leg images from 100 different individuals are captured using a Canon PowerShot A420 digital camera. The image is taken without any special lighting condition. The foot must be closed while acquiring the foot image. Dark intensity backgrounds, such as black and dark blue are used in this work. The foot image is saved using JPEG format in 1024 x 768 pixels.
Digital Camera
Footprint Image
Feature Extraction Matching Feature Extraction
Database
Accept or Reject
Fig. 2 Footprint identification using Feature Extraction
In this work, hundred images of left leg of 100 people are acquired using a digital camera. The foot image is aligned according to the keypoints and it is cropped. The energy
Fig. 3 Foot image
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
49
2.1.2. Image Cropping
2.1.4. LPCC and LPC
After acquiring the foot image, the next step is to crop the image. Cropping can be done by keypoints determination and extracting the image. The foot image is cropped to extract the heel shape as the intensity is highest in the heel portion. Fig. 4 shows the extracted footprint image using the proposed method.
The Grayscale image is further enhanced by converting to a matrix format of 256 X 256 pixels. Also the matrix so formed is separated into odd and even matrices to be used in the transformation technique.
Fig. 4 Footprint image
2.1.3. Conversion to Grayscale format The footprint image is acquired in RGB format. It is converted into grayscale intensity format before image enhancement. Fig. 5 shows the footprint image in grayscale format.
Fig. 5 Footprint image in Grayscale Intensity format
2.1.5. Transformation Techniques Footprint texture features are usually extracted using transform-based method such as Fourier Transform [5] and Discrete Cosine Transform [6]. While using Discrete Cosine Transform, some of the points are missed leading to an inaccurate inference. Also Fourier Transform involves floating-valued signals to integer-valued signals, thus less accuracy. Besides that, Wavelet Transform [7] is also used to extract the texture features of the footprint. In this work, a Sequential modified Haar Wavelet is proposed to find the Modified Haar Energy (MHE) feature. The Haar wavelet coefficients are represented using decimal numbers. It required eight bytes for storing each of the Haar coefficients. The division of the subtraction results in y-axis operation will also reduce the difference between two adjacent pixels. Due to these reasons, sequential Haar wavelet that maps an integer-valued pixel onto another Integer-valued pixel is suggested. Sequential Haar coefficient requires only two bytes to store each of the extracted coefficients. The cancellation of the division in subtraction results avoids the usage of decimal numbers while preserving the difference between two adjacent pixels. For every image of the detail coefficients, the image is further divided into 4 x 4 blocks. The Modified Haar Energy for each of the block is calculated using (1). PQ MHEi,j,k=∑∑(Cp,q)2 ------- (1) p=1q=1 where i is the level of decomposition, j is Horizontal, Vertical or Diagonal details, k is the block number from 1 to 16, P x Q is the size of the block. The MHE energy feature for every detail coefficients are arranged as in (2).
MHEi.j=[MHEi,j,1,MHEi,j,2,…..,HEi,j,16]--- (2)
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
50
2.1.6. Storing
2.3.2. Testing
Lastly, the Modified Haar Energy (MHE) or the threshold value of the footprint image is stored in the database for future references.
The footprint to be tested was taken in 10 different angles and heel part was cropped so as to get the MHE of the heel shape using:
2.2. Recognition Method
MHE=minimum (MHE1,………)
Recognition Method is also called as Testing. The normalization procedure removes the geometric information of input footprint (raw image). In Testing, normalization is done for the footprint image and then it is compared with the one stored in the database.
This test MHE was compared with the MHE of different persons stored in the database.
Result=TestMHE-VPersoni MHE
2.3. Experiments & Results
Where i=1 to 100
For experiment we took image samples of 100 different persons in 10 different angles using a digital camera. We used the block to get the conclusion.
If result is 0, the person with same footprint is found. That person is the master of the test footprint sample.
Comparison with other methods:
2.3.1. Training We took image samples of 100 different people and cropped heel portion to find the, MHE of all blocks. We took a number of MHE out of which we select the minimum. Let MHE1,MHE2, MHE3 …………….etc be the Modified Haar Energy for n blocks. Using the formula, MHE=Minimum (MHE1,………….) For each person, the MHE is found in this way.
VPerson1= Minimum(MHE1……………..) VPerson2= Minimum(MHE2……………..) ….. VPerson100= Minimum(MHE100…………) These are stored in the database.
The accuracy of Recognition of Footprint images using Sequential Modified Haar Transform is compared with the other methods for recognition. The following table shows the comparison of accuracy of recognition for Sequential Modified Haar Transform with Discrete Cosine Transform and Fourier Transform
TABLE 1: COMPARISION OF ACCURACY (PERFORMANCE) OF THE PROPOSED METHOD WITH OTHER WAVELET BASED METHOD
Transform Types
Recognition Accuracy (%)
Computation time in ms (Recognition)
DCT [Ref.6]
83.64 1
42
FT [Ref.5]
87.43 1
28
SHT [Proposed]
92.375 1
18
IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 3, No 5, May 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814
The corresponding graph for the accuracy rate of recognition footprints using above mentioned techniques in as follows
[5] Wenxin Li, David Zhang and Zhuoqun Xu, "Palmprint Identification By Fourier transform, ”Intl.Journal of Pattern Recognition and Artifical Intelligence, vol. 16 No.4, 2002 ,pp. 417-432. [6] Xiao-Yuan Jing and David Zhang, "A Face and Palmprint Recognition Approach Based on Discriminant DCT Feature Extraction,"IEEE Trans. on Systems, Man. and Cybernetics –Part B: Cybernetics, vol. 34 No. 6, Dec 2004, pp. 2405-2415.
94 92 90 A c c u r a c y (% )
51
88
[7] Xiang-Qian Wu, Kuan-Quan Wang and David Zhang,"Wavelet Based Palm print Recognition," Proceedings of the First Intl. Conference on Machine Learning and Cybernetics, Nov 2002, pp.1253-1257
86 84 82 80
Ambeth Kumar Devadoss was born in 1983. He
78 DCT
FT
SHT
Tranform types
Fig.6 Comparison of existing transforms (DCT, FT) Accuracy in % with proposed one (SMHET)
3. Conclusion
The heel portion of the leg is cropped as it is having more intensity at this portion. This cropping is done using built-in function. The heel portion is divided into blocks using Sequential Modified Haar Transform. Minimum MHE is selected from all the calculated MHEs. Comparing the MHE of test image with all person’s MHEs. If both MHEs match then the master of the footprint is found. This is an efficient method as it is more accurate. A high accuracy rate of 92.375% is achieved using Sequential Modified Haar Transform. This accuracy rate can be improved further by using Pressure sensing mat.
received the B.E. degree from the Department of Computer Science and Engineering, Madurai Kamaraj University. Madurai, India, in 2004.He received the M.E. degree from the Department of Computer Science and Engineering Annamalai University, Chidambaram, India, in 2006.He has published Four papers in International level Conferences and Five papers in National level Conferences .He has published one paper in National Journal. He is pursuing his PhD under the guidance of Dr.Ramakrishan .His research interests are in Image processing, Pattern recognition, Biometrics Dr.M.Ramakrishan was born in 1967. He is working as a professor and Head of PG department of Computer Science and Engineering in velammal Engineering College, Chennai. He is a guide for research scholars in many universities .His area of interest is Parallel Computing, Image Processing, Web Services and Network Security. He has 21 years of teaching experience and published 8 National and International journals and 40 National and International conferences. He is member of ISTE and senior member of IACSIT.
References
[1] “Special issue on automated biometrics, ” Proc\IEEE, vol. 85, pp.1348– 1491, 1997. [2] L. C. Jain et al., Ed., Intelligent Biometric Techniques in Fingerprint and Face Recognition. Boca Raton, FL: CRC Press, 1999. [3] R.Plamondon and G. Lorette, “Automatic signature verification and writer identification –The state of the art”, pattern Recog.,vol.22,pp.107131,1989. [4] Bimbot,H.P.Hutter,C.Jaboulet,J.Koolwaaij Lindberg, and J.B.pierrot,”Speaker Verification in the Telephone network: Research activities in the CAVE project,”in Proc.5th Int..Conf. Speech Communication and Technology, Rhodes, Greece, Sept.1997.
IJCSI CALL FOR PAPERS NOVEMBER 2010 ISSUE Volume 7, Issue 6 The topics suggested by this issue can be discussed in term of concepts, surveys, state of the art, research, standards, implementations, running experiments, applications, and industrial case studies. Authors are invited to submit complete unpublished papers, which are not under review in any other conference or journal in the following, but not limited to, topic areas. See authors guide for manuscript preparation and submission guidelines. Accepted papers will be published online and authors will be provided with printed copies and indexed by Google Scholar, Cornell’s University Library, ScientificCommons, CiteSeerX, Bielefeld Academic Search Engine (BASE), SCIRUS and more. Deadline: 30th September 2010 Notification: 31st October 2010 Revision: 10th November 2010 Online Publication: 30th November 2010
Evolutionary computation Industrial systems Evolutionary computation Autonomic and autonomous systems Bio-technologies Knowledge data systems Mobile and distance education Intelligent techniques, logics, and systems Knowledge processing Information technologies Internet and web technologies Digital information processing Cognitive science and knowledge agent-based systems Mobility and multimedia systems Systems performance Networking and telecommunications
Software development and deployment Knowledge virtualization Systems and networks on the chip Context-aware systems Networking technologies Security in network, systems, and applications Knowledge for global defense Information Systems [IS] IPv6 Today - Technology and deployment Modeling Optimization Complexity Natural Language Processing Speech Synthesis Data Mining
For more topics, please see http://www.ijcsi.org/call-for-papers.php All submitted papers will be judged based on their quality by the technical committee and reviewers. Papers that describe research and experimentation are encouraged. All paper submissions will be handled electronically and detailed instructions on submission procedure are available on IJCSI website (www.IJCSI.org). For more information, please visit the journal website (www.IJCSI.org)
© IJCSI PUBLICATION 2010 www.IJCSI.org
IJCSI The International Journal of Computer Science Issues (IJCSI) is a refereed journal for scientific papers dealing with any area of computer science research. The purpose of establishing the scientific journal is the assistance in development of science, fast operative publication and storage of materials and results of scientific researches and representation of the scientific conception of the society. It also provides a venue for researchers, students and professionals to submit ongoing research and developments in these areas. Authors are encouraged to contribute to the journal by submitting articles that illustrate new research results, projects, surveying works and industrial experiences that describe significant advances in field of computer science.
Indexing of IJCSI: 1. Google Scholar 2. Bielefeld Academic Search Engine (BASE) 3. CiteSeerX 4. SCIRUS 5. Docstoc 6. Scribd 7. Cornell’s University Library 8. SciRate 9. ScientificCommons
© IJCSI PUBLICATION www.IJCSI.org