138
Int. J. Mobile Communications, Vol. 3, No. 2, 2005
A framework for effective quality of service over wireless networks Jeremy Wee Deloitte, Wellington, New Zealand, 8 Cambridge Street, Tawa, Wellington, New Zealand E-mail:
[email protected]
Jairo A. Gutiérrez* Information Systems and Operations Management Department, University of Auckland, 7 Symonds Street, Auckland, New Zealand E-mail:
[email protected] *Corresponding author Abstract: This paper firstly highlights areas that have the potential to influence or directly affect the final outcome of Quality of Service (QoS) levels and the effectiveness of security in a wireless network environment. A Delphi methodology survey was then carried out by selecting a panel of experts with first-hand knowledge in this area. This research methodology process was conducted over three separate rounds of survey questionnaires for further refinement. The results gathered from the research indicated that the panel highly valued scenarios that supported further enhancements to existing QoS functions as well as continual development to newer and more cost effective functions that increase the level of control of QoS levels. Security was felt to be a secondary concern, a worrisome development given the increasing importance of it in modern networks (Siau and Shen, 2003; Olla and Patel, 2003; Tan, Wen and Gyires, 2003). Keywords: quality of service; wireless networks; wireless security; Delphi methodology. Reference to this paper should be made as follows: Wee, J. and Gutiérrez, J.A. (2005) ‘A framework for effective quality of service over wireless networks’, Int. J. Mobile Communications, Vol. 3, No. 2, pp.138–149. Biographical notes: Jeremy is an IT Consultant with Deloitte in Wellington (New Zealand). He received BCom (2001) and MCom (2003) degrees in Information Systems from the University of Auckland. Jairo is a Senior Lecturer in Information Systems at the University of Auckland. His current research topics are in network management systems, wireless networks, programmable networks, and Quality of Service issues associated with internet protocols. He received a Systems and Computer Engineering degree from The University of The Andes (Colombia, 1983), a Masters degree in Computer Science from Texas A&M University (1985), and a PhD (1997) in Information Systems from The University of Auckland (New Zealand).
Copyright © 2005 Inderscience Enterprises Ltd.
A framework for effective quality of service over wireless networks
1
139
Introduction
Data communication networks have been greatly enhanced since they were first introduced. More specifically, the media in which data travels is no longer restricted to the traditional form of copper wires that organisations used during the last few decades. Data signals can now be transmitted using radio wave technologies, which transmit wireless signals as electromagnetic waves that are capable of travelling through the vacuum of outer space and through media such as air. These kinds of data transmissions are often associated with the term wireless networks and they usually allow a sending node to be mobile as well as sustain the ability to carry out information exchange. Wireless networks often consist of mobile nodes that are interconnected by multi-hop communication paths. In other words, each mobile node has the ability to forward on data that is received from a sender onto another node that is most likely to be closer to the intended recipient on the other side. Conventional wireless networks often have a fixed network infrastructure and centralised administrative support for their operation. Apart from radio waves technology, several other technologies can also be used in conjunction when building a network. Other technologies include microwaves that provide another form of point-to-point connectivity as well satellite connection that are ideal when deploying wide area networks (WANs) that span across international boundaries. The structure of this paper is as follows: first, a literature review is conducted to discover the relevant research issues and group them into a three-level classification: Application, Middleware and Infrastructure. Secondly, the research methodology is presented and the research framework is introduced. Thirdly, a survey analysis is done on the results collected to highlight the potential areas that affect QoS in wireless networks. Lastly, a framework for effective QoS in wireless networks is introduced and discussed.
2
Literature review
Three categories are introduced in this section in order to systematically organise the various sources according to the area in which was felt to have the most significant impact in terms of QoS in wireless networks. These three categories are defined as follows: 1
Application: This category is concerned with issues linked to the software used to provide Quality of Service (QoS) in Wireless Networks and is not dependant on the underlying network infrastructure.
2
Middleware: This category deals with research that attempts to separate the network hardware infrastructure from the software and act as a liaison between the Application and Infrastructure categories.
3
Infrastructure: This category deals with the underlying wireless network infrastructure and is independent of the categories above it.
140
J. Wee and J.A. Gutiérrez
2.1 Application In radio networks, Lu (2000) felt that in order to support communication intensive real-time and non real time data flows over a scarce, varying and shared channel, there must be a service model that can react to the respective scenarios. Priggouris (2000) conducted further research on supporting IP QoS over the General Packet Radio Service (GPRS) and stated that the major deficiency of the current GPRS specification is the lack of adequate IP quality of service support. The authors also proposed solutions to the problem of establishing QoS reservations across the GPRS core network and the required signalling enhancements and modifications in the components of the GPRS architecture. The IETF-defined policy control architecture (shown in Figure 1) links high-level business requirements, such as those that can be specified in a Service Level Agreement (SLA), to low-level device implementation mechanisms, ranging from specific access control and management of services, objects and other resources to configuration of mechanisms necessary to provide a given service. The policy control architecture is made up of three types of entities: 1
Policy Repository
2
Policy Decision Point (PDP)
3
Policy Enforcement Point (PEP).
Figure 1
IETF policy control architecture
2.2 Middleware Campbell (1998) focused on developing a middleware called ‘Mobiware’, which is a QoS-aware middleware platform that handles the complexity of supporting multimedia applications operating over wireless and mobile networks. This middleware had the ability to control and manage all connections associated with a mobile device using a single connection group identifier (CGI), which uniquely represents a single reference point to manage all connections.
A framework for effective quality of service over wireless networks
141
2.3 Infrastructure Kim and Krunz (2000) conducted further research on analysing the packet-loss performance over a wireless link and consider cases of single and multiplexed traffic streams. By comparing the analytical results with realistic simulations, they observed that in the case of a single stream, the analytical expressions for the packet loss rate and wireless effective bandwidth are acceptable over a wide range of bit error rates. Partridge and Shepard (1997) on the other hand claimed that achieving these high data rates is difficult over satellite networks because you cannot get a TCP/IP implementation to perform well at higher speeds unless it supports large windows and speeds past 100 Mb/s.
3
Research methodology
This study utilises the Delphi Technique of surveying a panel of experts in the field of wireless networks. The primary purpose of the Delphi technique in this research is to obtain input from qualified individuals concerning problems or future directions or needs of quality of service in wireless networks. This process stops when consensus has been reached among the participants or when sufficient information exchange has been obtained (Dalkey, 1967).
3.1 Selection of participants This process involved selecting professionals from different sectors (vendors, users, researchers) of the network industry. The participants were selected from the listings contained within the IEEE New Zealand, Wireless Workshop 2001 seminar. Another set of participants were selected through referrals given by staff members within the University of Auckland. The range of companies represented in the survey included: •
Vodafone NZ
•
Telecom NZ
•
Nokia NZ
•
Unisys NZ
•
TelstraClear NZ.
Selections of these participants are not random. This is intentional for two reasons: 1
Delphi technique requires highly motivated participants in order to respond faithfully to several rounds of surveying.
2
All participants are experienced network engineers or analysts and therefore helped form an appropriate sample group for the purpose of the study.
3.2 Data collection process Three rounds or phases of surveys were used to collect data in response to a series of questions relating to the quality of service and security aspects of wireless networks. Round one involved the presentation of open-ended questions regarding what security issues were lacking, if any, as well as what the panel believed should be implemented in future designs involving security policies and mechanisms for quality of service in
142
J. Wee and J.A. Gutiérrez
wireless networks. The main goal of round one is to encourage participants to ‘collectively brainstorm’ ideas that can be further analysed in round two. Round two then requires this panel of experts to rate the identified and proposed issues so as to be able to further identify the more important issues out of the ones that were raised from round one. The final round requires the same panel of experts to rank their respective issues identified in round two according to a one-to-five point scale. The goal of this round is to verify with the respondents their selections made in round two and allow a final opportunity to make any final changes if required.
3.3 Analysis of results These findings were observed once all data was collated and analysed. It was observed that a majority of the QoS scenarios were ranked in the top five based on importance with the exception of scenario 7, which was more security oriented. Scenarios 1 and 2 attained similar ranking scores as the panel felt that it addressed the heart of the problem, that being inadequate QoS support in today’s wireless networks. The consensus that was achieved for these two scenarios was evident as there was a difference of 14 points between the top, second and third positions. The justification for this separation between the top two scenarios and third scenario appears to be the fact that current wireless networks are unable to support sustainable levels of QoS to users, especially when highbandwidth-demanding applications are required. Table 1 shows that many of the other security related scenarios were ranked closely behind scenario 7 indicating no real specific priority on any of the remaining scenarios. Figure 2 graphically depicts the overall results showing the general preference between the two factors. The percentages in Figure 2 are based on the total ranking scores that were collected from the third round. Table 1 Order of importance 1
Scenarios sorted by ranking score Scenario number 1
Scenario
Ranking score
Existing QoS adequacy
100 100
2
2
Future QoS adequacy
3
4
Optimal solutions
86
4
3
Call admission control schemes
84
5
7
Most apparent security threat
75
6
5
Business solutions
74
7
6
Spectrum resource constraints
74
8
11
Security protocol enhancements
61
9
8
Wireless security standard
58
10
9
Enforcement of security policies
54
11
12
Ideal security characteristics
54
12
10
Security mechanisms (wired/wireless)
37
13
13
Future user applications
24
A framework for effective quality of service over wireless networks Figure 2
143
Two factor comparisons: QoS vs. Security
2 Factor Comparison: QoS vs Security
Security 41%
Quality of Service 59% Security
Quality of Service
3.4 Analysis by QoS adequacy According to the results, it was noticed that in order for QoS to be considered adequate for high bandwidth-demanding applications that operate in a wireless environment, more attention should be placed on enhancing existing mechanisms as well as developing newer protocols and mechanisms as compared to achieving an optimal solution with the existing protocols and mechanisms. Figure 3
Security infrastructure comparisons
QoS Adequacy Call Admission Control Schemes 23%
Optimal Solutions 23%
Existing QoS Adequacy 27%
Future QoS Adequacy 27%
Existing QoS Adequacy
Future QoS Adequacy
Optimal Solutions
Call Admission Control Schemes
144
J. Wee and J.A. Gutiérrez
3.5 Analysis by security infrastructure Despite the lack of priority given to the security scenarios, one of the security related scenarios stood out above the rest. The panel felt that the most important issue that should be addressed in terms of wireless networks is the threat of passive eavesdropping, which makes up for 33% of the total score in this category. An interesting point to take note of here is that even though the results are within the scope of wireless networks, many experts expressed a wider applicability when referring to such security threats. In other words, passive eavesdropping and the likes of it is an apparent threat in both wired and wireless networks and should therefore not be typecast as a threat belonging only to the latter. Figure 4
Security infrastructure comparisons
Security Infrastructure
Security Mechanisms 16%
Ideal Security Characteristics 24%
4
Most Apparent Security Threat 33%
Security Protocol Enhancements 27%
Most Apparent Security Threat
Security Protocol Enhancements
Ideal Security Characteristics
Security Mechanisms
Framework formulation
With the information gathered from the study combined with the information gathered from the literature review, a framework is proposed which hopes to assist vendors and service providers in constructing an effective QoS-enabled wireless wide area network. This framework comprises of multiple components that actively interact with one another whilst systematically addressing each layer of the wireless network. The framework is essentially split up into three layers, which are similar to the categorisation used in the literature review section. The Application, Middleware and Infrastructure layer provides a systematic structure to the framework as well as defines the scope of the components within the framework.
A framework for effective quality of service over wireless networks
145
4.1 Global security boundary This global security boundary encapsulates all components in the framework indicating that all tasks that are executed in the network are done in a secure environment in a standardised fashion. In other words, individual networks are provided with the flexibility of implementing higher levels of security measures but must however maintain a certain minimum level of security as specified by the global information model. Each individual network however does not possess the ability to extend the security boundary and would require the authorisation of higher-level management before doing so.
4.2 QoS global control mechanism The QoS global control mechanism involves the governing of the mechanisms used throughout the entire wide area network. Therefore all QoS related policies as well as QoS mechanisms are managed via this component. The QoS control mechanisms receive the relevant information on the different services and devices from the information model, which are utilised in the network and determines the level of enforcement required for each service. Using this same information from the globally defined information model, further information is retrieved from the policy repository, which houses all policy types that have been administered both in the past and present.
4.3 Policy repository The administrator typically defines the information contained in the policy repository. These policies are largely high-level policies that govern the wide area network. The scope of the policy rules will vary to a certain extent. Global policies have a global effect on the end result, meaning that the end result of the global policy relies on being executed globally across the entire path of the network service (Gomez, 2002). In this repository, there also exist domain policies as well as local policies that govern specific networks. The domain policies are meaningful within one domain of QoS mechanisms, for example radio networks whilst local policies are relevant only to one network element. This component remains consistent with the policy architecture that has been proposed by the Internet Engineering Task Force (IETF) and which was mentioned in Section 2 of this paper.
4.4 Service applications layer Similar in characteristics to the above components, this application layer purely serves as a service application library that is extendable if required. If the QoS global control mechanism defines the policies and mechanisms for a new service application, the concerned administrator would thus have to ensure that the application is available when being requested by users. The advantage of having the service application layer is to govern the usage of the type of service application that is requested and append the corresponding defined policy and mechanism that is specified by the global QoS control mechanism. This establishes a central control point for the entire network, which facilitates an easier control over the applications used on the network as well as their individual usage.
146 Figure 5
J. Wee and J.A. Gutiérrez Proposed framework for effective QoS over wireless networks
A framework for effective quality of service over wireless networks
147
4.5 Application Program Interface (API) This API layer performs its typical function of splitting the applications into building blocks, which will allow integration with other applications. Similar to the APIs that are available in the Windows Operating System, this layer converts all service application attributes to DiffServ code-points so that they are understandable by the DiffServ component. A probable candidate that can execute such a task is the model proposed by Xiao (2000). This model was felt to be suitable as its primary goal is to enhance and utilise the DiffServ standard for mobile ad-hoc networks. A potential downfall of utilising this model here however is that the model was built to sustain a flat hierarchical topology of 50 nodes, which would therefore be ineffective once the network expands beyond this limit.
4.6 Specific QoS required functions This component allows more flexibility to be built into the network if necessary. At this point, the respective administrator will decide on how many more extensions are required to suit the individual needs of each different local area network. Seeing as different LANs will utilise different applications and services from other LANs within the same WAN, it is important that the administrator impose further QoS controls that will enhance the existing controls that are defined by the global QoS control mechanism. Imposing such a control remains consistent with the findings from the Delphi methodology as expert two, for instance, indicated “The optimum solution for one network could be irrelevant to another so it is unlikely there will be a universal optimum solution.”
4.7 QoS enhanced functions This component adds further flexibility to the proposed framework as it allows administrators and developers to consider newer or more enhanced QoS functions that can be integrated with the existing QoS functions. This is illustrated with a dotted line boundary as it presents itself as a window of opportunity for newer technology to be integrated into the framework. This is illustrated by coupling this component with the specific QoS functions detailed above.
4.8 Specific security required functions Similar to the specific QoS required functions; this component requires that the concerning administrators or developers extend the global security infrastructure that is already in place. Therefore if certain networks require the services of a Virtual Private Network (VPN) then extra tunnelling protocols will need to be implemented here. This coupled with enhanced authentication schemes provides a higher level of security measures between networks while still utilising the base security functions that are defined in the global security policy.
148
J. Wee and J.A. Gutiérrez
4.9 Security enhanced protocols/mechanisms This component is also largely similar to the QoS Enhanced Function component as it once again allows administrators and developers to consider newer or enhanced security protocols or mechanisms that can be integrated with the existing security measures taken. Once again it is illustrated as a dotted line boundary as it presents itself as a window of opportunity for newer technology to be integrated into the framework. Evidence collected from the security protocol enhancement scenario suggests that the panel largely supported the possibility of future enhancements and expansion of future QoS security enabled wireless networks.
4.10 Local security boundary Similar to the globally defined security boundary, the local security boundary outlines the region that proprietary solutions will cover between the different network infrastructures. This assumes that the network administrator has ventured into proprietary solutions, which can be integrated via APIs. Situated within each local security boundary would be local policy management components, which enforce globally defined QoS protocols as well as individually extended security protocols and mechanisms. The local management component will then deploy both global and local QoS and security policies to the individual Policy Enforcement Points (PEP).
4.11 Feedback mechanism This proposed framework suggests a largely controlled environment with central control points that deploy pre-determined definitions further down the layers. The feedback mechanism allows for more accurate results to be fed back to the globally defined QoS control and the information model based on the currently deployed policies and mechanisms. Should any discrepancies exist within the deployment of all policies and mechanisms, all individual network infrastructures will possess the ability to send periodic update reports to the information model and global QoS control repository for further refinement and eventually more effective deployments to the different devices on the network. The feedback mechanism spans between the secured and unsecured regions of the framework indicating the nature of the feedback itself. Feedback that is destined for the information model should largely contain abstract level type information whilst any feedback to the global QoS control component should be of a certain degree of detail outlining possible improvements based on the more unstable sections of the network.
4.12 Framework limitations A limitation that is rather apparent in this framework is the lack of complete flexibility as there are only designated windows where new technologies can be integrated into the framework. The majority of the framework relies on the environment being largely hierarchical therefore a certain level of compromise was required in order to achieve the goal of providing effective QoS control over wireless networks.
A framework for effective quality of service over wireless networks
5
149
Conclusion
Based on the evidence collected from this research, the majority of the panel expressed brighter prospects for future wireless networks despite its current inadequacies in providing sufficient resources for bandwidth-intensive applications. The panel also largely agreed that the success of future wireless networks would not be purely based on the performance of a single application but rather on how cost effective the solution would be as a whole. Commercial viability was thus found to be the main driving force in determining the success or failure even though the implementation of a QoS enabled wireless network might already be technologically feasible. The findings from the data analysis helped shape the framework to be more QoS oriented whilst still incorporating a certain degree of security related characteristics. It was important to allow a certain degree of flexibility into the framework as this technology is still largely in its development stages. Several areas within the proposed framework allow for some integration and reconfiguration (an increasingly popular requirement for wireless networks, see Panagiotakis, et al.) of new and current technologies to occur without having to completely restructure the components found at the lower layers of the framework.
References Campbell, A.T. (1998) QoS-Aware Middleware for Mobile Multimedia Communications, New York: Columbia University, pp.63–78. Dalkey, N. (1967) Delphi, Rand Corporation. Gomez, G. (2002) ‘QoS policy management in 3G mobile networks’, Advance: Nokia Technology Magazine, pp.41–43. Kim, J. and Krunz, M. (2000) ‘Bandwidth allocation in wireless networks with guaranteed packet loss performance’, Communications of The ACM, Vol. 8, No. 3, pp.337–349. Lu, S. (2000) ‘Design and analysis of an algorithm for fair service in error-prone wireless channels’, Wireless Networks, Vol. 6, pp.323–343. Olla, P. and Patel N. (2003) ‘A framework for delivering secure mobile location information’, International Journal of Mobile Communications, Vol. 1, No. 3, pp.289–300. Panagiotakis, S., Koutsopoulou, M., Alonistioti, A., Houssos, N., Gazis, V. and Merakos, L. (2003) ‘An advanced service provision framework for reconfigurable mobile networks, International Journal of Mobile Communications, Vol. 1, No. 4, pp.425–438. Partridge, T. and Shepard, C. (1997) ‘TCP/IP performance over satellite links’, IEEE Network, September–October, pp.44–49. Priggouris, G. (2000) ‘Supporting IP QoS in the general packet radio service’, IEEE Communications Magazine, September–October, pp.8–17. Siau, K. and Shen, Z. (2003) ‘Mobile communications and mobile services’, International Journal of Mobile Communications, Vol. 1, Nos. 1–2, pp.3–14. Tan, J., Wen, H. and Gyires, T. (2003) ‘M-commerce security: the impact of wireless application protocol (WAP) security services on e-business and e-health solutions’, International Journal of Mobile Communications, Vol. 1, No. 4, pp.409–424. Xiao, H. (2000) ‘A flexible quality of service model for mobile ad-hoc networks’, Proceedings of The Vehicular Technology Conference, 2000, VTC 2000-Spring, Tokyo, May, pp.445–449.
