a distributed workflow management system utilizing computational ...

A DISTRIBUTED WORKFLOW MANAGEMENT SYSTEM UTILIZING COMPUTATIONAL INTELLEGIENCE FOR DECISION SUPPORT Z. J. Muhsin

A. H. El-Mousa

M. A. Al-Taee

Computer Engineering Department, University of Jordan, Amman 11942, Jordan e-mail: [email protected], [email protected], [email protected]

ABSTRACT Distributed workflow management systems are increasingly gaining momentum due to the emergence of virtual and distributed organizations, as well as cross-organizational workflows. This paper presents an effective development process for a distributed workflow management system (WKFMS) which is utilized to automate the management processes related to computer and peripheral repair requests in the Faculty of Engineering and Technology (FET) at the University of Jordan. First, the paper explores the reasoning and motivation behind the choice of the management processes involved in the WKFMS. Phases of the development process that is based on the rapid prototyping paradigm are then described. Particular focus is given to the modeling of business scenarios using the Model-View-Control (MVC) architecture. The distributed network layout and system security measures are also investigated. Reduction of manual human intervention in the decision process is achieved through the implementation of a neuro-fuzzy computational intelligence processor to automate the major decision making process of tasks distribution. It is shown that the application of the proposed development process helped in reducing the overall system development time and effort. The successful deployment and functional testing of the developed industrial prototype has proved that it works smoothly according to the desired features. Index Terms — Web-based distributed systems, workflow management, software prototyping, network security, computational intelligence.

1. INTRODUCTION In today’s information technology society, it is no longer acceptable nor is it cost conscious to keep to the old paperbased approaches in managing business processes. This is even more seen as being totally ineffective, if the infrastructure is already there to use electronic based transactions. Therefore, it is not surprising that from early on, serious efforts were directed at utilizing whatever technologies were available at the time, to automate repetitive tasks using IT [1]. In the field of equipment management, many efforts were attempted. A multi-user computer system for the management of hospital equipment including repair requests, inventory control, repair costs and service records was described by Whelpton and Cooke [2]. The system uses a relational database together with

personal computers. However, data entry transferred from paper requests is done by only one user and there is no methodology to follow up requests or monitor activities. A similar system that manages information used for the management of preventive maintenance and equipment repairs at the University of California Berkeley was also described by Mudie and Chang [3]. The system provides technicians and management with a tool for monitoring equipment maintenance events, reviewing equipment history, and resolving equipment malfunctions. The system allows equipment users to report faults which are automatically forwarded by email to maintenance staff. A relational database is used to store information regarding events. The system lacks online direct monitoring of events and status. Also, it delegates the management of all events to the maintenance staff which reduces accountability. Recently, it has been shown that projects developed using traditional software lifecycle models such as the waterfall model typically fail for several reasons, namely: output-driven orientation of design process, application backlog due to the document maintenance workload and user dissatisfaction [4] [5]. For small and medium sized systems, a development methodology based on rapid prototyping approach is intended to correct these shortcomings and to bring a software project closer to its user(s). After several iterations of the prototype, the developers have a better understanding of the requirements and the users have a better idea about how the system will eventually look and feel. For large systems, the prototype based methods have difficulty in identifying the global aspects and boundaries of the project and design inconsistencies and shortcomings [5, 6]. However, despite their difficulties in large system developments, prototype-based methods have been a hot topic in software engineering for more than a decade. Different design models for workflow management systems have been reported in literature, using: Petri Nets [7], agent technology [8], knowledge-based [9], document-based [10] and more recently object-oriented modelling. When compared to other modelling methods, the object-oriented modelling approach offers improved flexibility and adaptability [11]. In the present paper, the development process, business process modules, network layout and system security of a distributed Workflow Management System (WKFMS) are presented and explored. Also, through the study of the workflow system and its working features, a neuro-fuzzy computational intelligence processor is suggested to automate the decision making process of tasks distribution. The system structure for this processor is presented and investigated. The static and dynamic software design models and development technologies of this WKFMS were previously reported in reference [12].

2. PROBLEM STATEMENT The existing work flow management system for computer repair and maintenance at FET involves a conventional paper-based process for order management. This process deals with order submission and workflow control activities as shown in Figure 1. The choice to transform the paper-based repair system business processes to an electronic form was first evaluated for suitability, based on the following factors: • The processes involved required people to be notified of events as they occurred (request progress). • The processes were approval-oriented processes (there is a management and control hierarchy in FET). • The processes cross functional and/or department lines (repair requests are directed from departmental or managerial units to an independent centralized repair and follow-up section). • Processes where workflow will add value to process stakeholders (expected reduction in time, effort and improved supervision and workload distribution). • Automation of processes contributes towards university goals (there is a clear direction towards process automation using modern IT structures and policies in the University of Jordan). • Automation has potential for significant improvements (requests can be remotely monitored, and repair workers are directly accountable. Also, the system has the potential to grow and interlink with the wider university IT system).

Figure 1. The existing paper-based order management process.

• Processes are feasible to redesign (since the processes involved already are form-based with a clear flow, then it is totally feasible).

3. SYSTEM OVERVIEW In addition to re-engineering the order management process mentioned above, the proposed system is supported by several newly added processes, namely: user management, report generation, real-time end-user follow-up, and a configurable management path. Figure 2 overviews the proposed distributed WKFMS for computer repairs and maintenance in FET at the University of Jordan. The three-tiered architecture is considered appropriate for the proposed WKFMS [12]. The tiers of the proposed architecture can be seen as having clients at the front end, services at the middle tier, and data storage at the back end. Recent advances in Internet technology and the availability of suitable software environments offer the possibility of using different implementation strategies and tools. Figure 3 shows the architecture based on a Java 2 Enterprise Edition (J2EE) platform which is proposed for the Workflow Management Information System (WKFMS) under consideration. When a web browser is used as the client interface, J2EE specifies the use of Servlets and JavaServer Pages (JSP) technology to generate web pages. JavaBeans are used to encapsulate business logic and to connect to the data tier through drivers based on a Java Database Connectivity (JDBC) standard. In the present work, the Relational DBMS (RDBMS) is considered in response to the constraints specified by the client. Moreover, the process of message exchanges between the application server and system users is performed through an existing Simple Message Transfer Protocol (SMTP).

Figure 2. Overview of the proposed WKFMS. Features of the proposed workflow management system can be summarized as follows: • Submission, distribution and monitoring of services are completely performed over the web. • Secure access levels to system features, depending upon the different categories of system users. • Fast generation of reports for past and present activities. • Real-time follow-up of work progress and timely notification of status. • Optimising job allocation and monitoring of personnel performance over a defined period of time. • Includes options for the job approval process that suits different styles of management. • Supports both Arabic and English languages.

4.

DEVELOPMENT PROCESS

In the prototyping paradigm, functional specifications are jointly defined by the system developer and users who are encouraged to revise and change their requirements [13]. The design phase which leads to the construction of a prototype, focuses on a representation of aspects that will be visible to the end user. The developed prototype is then deployed on the client’s environment and evaluated by the system end users.

Figure 3. Interaction between system development phases. Iteration occurs as the prototype is tuned to satisfy the needs of the user. Feedback from system users is used by the developer to better understand what needs to be done and thus is used in refining the requirements of the software. A major decision has to be made, when following a prototyping paradigm, is whether to build a full- sized production prototype or a scaled down model. The full-scale prototype is considered more appropriate for the proposed workflow management system, due to the following reasons: the system functionality is fairly well understood, the number of functions is clearly defined and the development tools are available. The development process adopted in this work is started with an envisioning and requirement analysis phase where the system is divided into several subsystems. The process then proceeds with three iterative phases related to the subsystem’s development; design, programming and evaluation. Each subsystem cycle is completed in a period of one to three weeks, as illustrated in Figure 3. Finally, the system development process is completed by integrating the developed subsystems and evaluating the complete system prototype.

5.1. User Registration and Management User registration starts by filling in the required information in a registration form. Once completed, the registration form is sent to the corresponding department manager for approval. Upon approval, the system automatically updates the user status, notifies the FET user via e-mail and sends the registration form to the administrator. The administrator then activates the user by updating the user’s status in the database. Upon user activation, the system automatically notifies the newly registered user via e-mail. Other users (i.e. the corresponding departmental manager, lab. manager and the FET Dean) can also be notified depending on the configurations of the corresponding contacts profile. Figure 4 shows the MVC model of the user registration and management process.

5. MVC MODELS OF PROCESS SCENARIOS Process scenarios of the WKFMIS are described briefly in this section. These scenarios include four processes: user registration and management, order submission and management, e-mail notification and reporting. Of these, the latter two processes are newly added processes to the existing paper-based system. These scenarios are based on observations and information gathered from users in the working environment in FET. The Model-View-Control (MVC) model (or architecture) which is widely used for interactive applications is also presented in this section. This architecture divides the functionality among objects involved in maintaining and presenting data to minimize the degree of coupling between the objects [14].

Figure 4. MVC model of the user registration and management.

5.2. Order Submission and Management Similar to the user registration and management process, the order submission and management process starts by filling in an order form. Once completed, the order form is sent to the corresponding departmental manager for approval. Upon approval, the system automatically updates the order status, notifies the FET user via e-mail and sends the order form to the lab manager. The lab manager then specifies a target completion time (in days) and forwards the order to one of the lab workers. The system also notifies the FET user, departmental manager and the assigned lab worker via e-mail. The lab worker will then perform the required service and update the order status according to the work progress until completion. For each order status update (e.g., started, pending, overdue or completed), the system sends email messages to the corresponding users. Submitted orders can also be cancelled by the users who have ordered them. The MVC architecture of the order submission and management process is shown in Figure 5.

Figure 5. MVC model of the order submission.

Table 1. Events of e-mail notifications for system users. New Usr New submitted order New order is approved by dept mger New order is approved by lab mger New order is started by lab worker Order is completed Order is overdue Active order cancelled by FET user Active order cancelled by dept mger Active order cancelled by lab mger New user is registered New user approved by dept mger User is inactivated by adm. Assigned to a new order

FET Usr

Dept. Mger

Lab Mger

Lab Wkr

Sys. Srvr

* * * * * * * * * * * * *

5.3. E-mail Notification This process aims at generating automatic e-mail notifications to all system users depending upon certain process events related to each user type. Table 1 summarises the events for which system users receive email notifications. The actor that initiates each event is designated with symbol (*) and the users to whom the messages are sent, are identified by shaded rectangles, as illustrated. Each user configures his/her contact profile via checkboxes available on the contacts page of each user. The MVC model of the e-mail notification (contact) process is shown in Figure 6.

5.4. Reporting Process This process aims at preparing and printing various types of reports. The contents of these reports vary depending on the user type. For example, FET users are allowed to prepare and print reports related to their own orders while the FET Dean can prepare reports on all orders and users existing in the system database. The contents of these reports can be filtered by status, department and time period. Table 2 summarises the types of reports which can be prepared by each user type. Unlike other processes, this process only retrieves, views and prints orderrelated and user-related information existing in the system database. As there is no need for complex processing in this process, the JSPs of all users can be connected directly to the database via a database connection class (DbConnection), as shown in the MVC model of Figure 7.

Adm.

Figure 6. MVC model of the e-mail notifications.

Table 2. Types of users reports. FET User

Dept Lab Lab Mger Mger Wkr

Admn.

Dean/Vice Dean

using this technique proved encouraging. However, due to space and time restrictions, it is not possible to elaborate at this stage. Figure 8 shows the modules utilized in such an implementation.

Personal orders reports Dept. orders reports FET orders reports FET users reports Figure 8. Neuro-Fuzzy decision support system. The possibility of using data mining techniques to provide the end users with automatic solutions for well known problem is also being explored. The system will utilize a database engine to store a description of the problem submitted by the end user through a from-based input process and the solution that was used to solve the problem. This engine will be utilized to provide a possible solution to similar new user submitted problems automatically without having the task followed-up online by the FET-lab workers.

7. FET LAB NETWORK AND SECURITY

Figure 7. MVC model of the reports process.

6.

AUTOMATION OF BUSINESS DECISION

In order to streamline and automate the workflow as much as possible, several issues are being explored that utilize computational intelligence techniques. It is obvious that among the major factors which impact system performance in the workflow model, are the processes which require direct human intervention. If the responsible person does not act quickly on job requests, this will slow down the system considerably. An important example is the task of assigning and distributing job requests by the lab manager to lab workers for action. The decisions taken by the lab manager are manual and they are at his\her discretion [12]. The lab manager will depend on experience and previous information regarding current allocated jobs to assign tasks in a balanced manner. To automate this process such that the system will take on this responsibility directly, the use of a neuro-fuzzy computational intelligence processor is investigated. Selection of a neuro-fuzzy approach was motivated by several aspects; the fuzzy logic part easily mimics the decision support process handled by a live lab manger. The Artificial Neural network is used in a continuous learning mode to update the parameters of the fuzzy decision support system to improve its results. Our initial investigation

In such a distributed WKFMS system, where there are forms to be submitted and an electronic approval process, security measures become essential to prevent unauthorized access and system misuse by both internal and external sources. In early 2006 the FBI in the USA estimated that computer crime was costing U.S. businesses $678 billion per year [15]. Application breaches can affect web-based systems through the loss of some or all of the following security goals: confidentiality, integrity; and availability [16]. The countermeasures that deal with these problems involve: access control, flow control, inference control, and encryption [17]. Due to space limitations, only the major aspects of these measures will be considered here. The measures taken in our system to insure security and privacy are the following: a) The inclusion of an ISA server acting a firewall between the servers and FET users as shown in figure 9. This server implements a specific security policy to prevent unauthorized access based on local IP addresses and uses a three-legged configuration setting. A DMZ was not considered here since FET is not directly connected to the Internet. b) User authentication. Similar to many applications, a user authenticates himself/herself by entering a user login ID and a secret password known solely to the user and the system. The security credentials of the user are encrypted and kept in database and they are never displayed to public in any way. c) User access privilege levels. Access privileges are manipulated in the users’ records depending upon user types. Each user is classified with one of the six roles: dean, administrator, department manager, lab manager, lab worker or FET user. Each user type can only access information records either belonging to himself (or herself) or belonging to other users in his/her privileged capacity. The above measures have been found to be adequate to insure system integrity, continued functionality and privacy.

direct solutions to known problems taken from the system database, thus allowing the end users to solve their issues, if possible, without having to register a repair request. The generation of automated daily work schedules for lab workers is also being considered to streamline the workload process. 10. REFERENCES

Figure 9. FET lab network connections to FET users.

8. RESULTS AND DISCUSSIONS The system is currently being implemented and ready for full deployment. The obtained results indicate that several immediate improvements have become apparent. Table 3 shows extrapolated estimates for performance compared with the original paper-based system. Table 3. Comparison between the new and old WKFMS. Process Activity Place Order Approve Order Distribute Order Trace Work Progress Annual Order/User Report

Estimated Processing Time Proposed system 2 min 0 – 1 min

Paper-based System 5 min 15 min – 4 hours

0 – 2 min

15 min – 1 hour

2 min

14 min – 4 hours

5 min

1 – 2 days

More than 25 connected pages (JSPs) are designed and implemented to display various types of information. These pages are categorized into six groups: common pages used by most or all users, FET user pages, department manager pages, lab manager pages, lab worker pages and administrator pages. For brevity, however, only samples of these pages are presented in Table 4 with brief comments on each page.

9. CONCLUSIONS A distributed workflow management system is presented for the management, follow-up and status update of user requests for computer repair and maintenance at the University of Jordan. The detailed development process, MVC architectures of management processes, models and security measures are shown along with the motivation and benefits of the developed WKMS. It is seen that the proposed full-scale prototyping development approach is an effective approach in maintaining direct and constant interaction between system developers, users and administrators. A neuro-fuzzy computational intelligence processor is suggested to automate the decision making process of tasks distribution, thus letting the system automatically assigns jobs as they arrive. The initial successful indications obtained from such implementation allow easy expansion in the utilization of artificial intelligence techniques in all other decision making aspects of the WKFMS under consideration. Also, we are considering introducing extra processes, also based on computational intelligence approaches, to give end users

[1] C.A. Ellis and G.J. Nutt,” Office Information Systems and Computer Science”, Computing Surveys, Vol. 12, No.1, pp. 27-60, 1980 [2] D. Whelpton and D.K Cooke,”Computer system for equipment management”, Journal of Biomedical Engineering, Vol. 12, No. 3, pp. 248-252, 1990. [3] D.C. Mudie and N.H. Chang, “FAULTS: An Equipment Maintenance and Repair System Using a Relational Database”, Proceedings of the IEEE/CHMT ‘90 IEMT Symposium, pp. 41-46, 1990. [4] D. E. Avison and G. Fitzgerald,” Where now for development methodologies?”, Communications of the ACM, Vol. 46, No. 1, pp. 78 – 82, 2003. [5] G. Milosavljevic and B. Perisic,”Really rapid prototyping of large-scale business information systems”, Proceedings 14th IEEE International Workshop on Rapid Systems Prototyping, pp. 100 – 106, 2003. [6] A. Krabbel , I. Wetzel, and H. Zullighoven,” On the Inevitable Intertwining of Analysis and Design: Developing Systems for Complex Cooperation”, Symposium on Designing Interactive Systems, pp. 205-213, 1997. [7] K. Salimifard and M. Wright, “Petri net-based modeling of workflow systems: An overview”, European Journal of Operational Research, Vol. 134, No. 3, pp. 664-676, 2001. [8] M. Crowe and S. Kydd, ”Agents and suggestions in a Webbased dynamic workflow model”, Automation in Construction, Vol. 10, No. 5, pp. 639 – 643, 2001. [9] S. Bin and Y. GuoFu,” The modeling method of workflow management system for agile enterprise”, Computer Integrated Manufacture System (CIMS), Vol. 8, No. 8, pp. 105-110, 2002 [10] J. Kim, W. Suh, and H. Lee, ”Document-based workflow modeling: a case-based reasoning approach”, Expert Systems with Applications, Vol. 23, No. 2, pp. 77-93, 2002. [11] L. Lin, S. Zhong and Yu. Yao,”Object-oriented distributed workflow management system”, Proceedings of the Ninth International Conference on Computer Supported Cooperative Work in Design, Vol. 2, pp. 685 – 689, 2005. [12] Z. J. Muhsin, A. H. El-Mousa and I. Potts, “Development of a Web Based Database-Centric Workflow Management System Using Rapid Prototyping Approach”, Proceedings of IEEE/ACS International Conference on Computer Systems and Applications, pp. 92 – 99, 2007. [13] F. Davis and V. Venkatesh, “Toward pre-prototype user acceptance testing of new information system: implications for software project management”, IEEE Transactions on Engineering Management, 51(1), 2004. [14] R. S. Pressman, Software Engineering: A practitioner’s Approach, McGraw-Hill, USA, 2005. [15] M. Shacklett, “IP Address-Level Security a Growing Concern”, Enterprise Networks & Servers Journal, November 2006. [16] E. Bertino and R. Sandhu, “Database security – concepts, approaches, and challenges”, IEEE Trans. on Dependable and Secure Computing, 2(1), 2005, pp. 2-19. [17] R. El Masri and S. B. Navathe, “Fundamentals of database systems”, Addison-Wesley, 2003.