Vertical Handover Decision Policy based on the End User's Perceived Quality of. Service. Sassi Maaloul, Mériem Afif, Sami Tabbane. MEDIATRON: Research ...
2013 27th International Conference on Advanced Information Networking and Applications Workshops
Vertical Handover Decision Policy based on the End User’s Perceived Quality of Service
Sassi Maaloul, Mériem Afif, Sami Tabbane MEDIATRON: Research Unit on Radio Communication and Multimedia Networks Higher School of communication of Tunisia (SupCom) Carthage University Tunis, Tunisia {maaloulsassi, mariem.afif, sami.tabbane}@supcom.rnu.tn stages: initiation process, network selection and the handover execution. The most important issue is how to select dynamically the best access network for mobile’s user which can be used and pursue its application services. As a result of the diversification of the context-aware and the impact of the perceived quality on mobile user, most of the existent handover mechanisms are not sufficient any more to make decision. There is necessity to develop a novel method that reduce handover latency and check to improve the quality of service perceived by the end users. The remainder of this paper is organized as follows. Section II gives an overview of the related works. In section III we define and present the context-aware used in our work. Section IV describes the Analytic Hierarchy Process (AHP) for weighting the context criteria. Section V provides a detailed overview of some related work as Simple Additive Weighting (SAW) and Weight Product Method (WPM) methods for ranking the access network. Section VI present our proposal method for network selection. Numerical analysis is given in section VII with evaluation of the described solution. On an ending note, the last section will conclude our work with some observations and results.
Abstract—The integration of heterogeneous wireless networks allows the users to benefit simultaneously from these radio access technology (RAT) and they can also use an important number of applications. New applications constraints are more complex and may be change dynamically and rapidly under time. The handover mechanism is responsible to guarantee the required quality of service when users roam across different RAT in order to provide an ubiquitous environment. With the diversification of the context metrics allowed by these networks, the level of computationally and complexity has a direct impact on the perceived quality of service especially of the network selection process on application performance. The present work addresses the network selection process, which is probably the most important key of the vertical handover decision mechanism. It proposes a modification of the Simple Additive Weighting called Enhanced-Simple Additive Weighting (E-SAW) for ranking networks that avoid computational cheaper by eliminating network access which does not satisfy a minimum required context. After ranking, in order to avoid bad network selection which does not respond to user requirement, we define handover decision scheme that improves network performances perceived by end users and avoid the processing delay caused by unnecessary handover.
II.
Keywords-Vertical Handover Decision Making; ContextAwareness; Analytic Hierarchy Process; Multiple Attribute Decision Making; Perceived Quality of Service
I.
In the recent years, most proposals were enriching the literature of the interoperability between heterogeneous wireless networks. Many researchers have been proposed that use context information to make handover decision. It’s addressed to the optimization of the context-aware by the use of different optimization policy. For example, the authors in [1] give an overview of the most interesting and recent algorithm to make vertical handover decision. They classify such algorithms in different categories to compare each algorithm to the others: Traditional “RSS-based” algorithm, Function-based algorithm, User-centric algorithm, Fuzzy Logic and Neural Networks algorithm, Context-aware strategies and Multiple Attribute Decision Making (MADM) (based on: SAW, Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), AHP and Grey Relational Analysis (GRA) algorithm. Authors in [2], propose a context-awareness handover based on a planning mechanism in heterogeneous wireless networks. They develop two integrated approaches for context-awareness handover planning mechanisms: AHP for assigning the weight of each level and TOPSIS for making decision approach and Genetic Algorithm (GA) approach which
INTRODUCTION
The mobile telecommunications technologies future generation is progressing steadily. It can offer various services to mobile users, in order to ensure good network performances and to provide ubiquitous broadband access. The interconnection and interoperability between these heterogeneous wireless networks is a critical challenge when users roam across different radio access networks. The major high-level requirements must be achieved by the next generation is Providing better Quality of Service (QoS) to end users. Therefore, the consideration of context awareness in heterogeneous wireless networks will be more adequate by offering to users different information of networks availability. It can change dynamically and it’s expected to meet continuously the increase of the application requirements. The handover procedure between different infrastructures should be based on politics of gathering information provided by the networks and required by mobile users. However handover process includes three main 978-0-7695-4952-1/13 $26.00 © 2013 IEEE DOI 10.1109/WAINA.2013.127
RELATED WORK
493
guarantees QoS requirements. An algorithm for a contextaware network selection is proposed in [3], it is based on a modified WPM for access network selection. It uses a weight distribution method to assign the weight of each criterion and make decision based on WPM and TOPSIS method. Also in [4] authors present an overview of vertical handover techniques during different stages: Handover information gathering, Handover decision and Handover execution. They proposed and classified the main vertical handover decision algorithm used in the literature (AHP, GRA, TOPSIS, SAW…) into different category depending on diverse metrics and parameters to evaluate the best candidate network in order to make the adequate decision. In [5], authors present a comprehensive survey of Vertical Handover Decision (VHD) algorithms designed to satisfy and to provide the required QoS. They describe the algorithms based on the main handover decision criterion used and evaluate tradeoffs between their implementation complexity and efficiency. The authors of [6] compare SAW and WPM methods to select the best network. They gave an overview of reducing the processing delay in handover mechanism. There are others work summary to improve handover decision making in heterogeneous system [7-11]. The particularity of these methods is the use of different MADM for weighting and ranking all network detected. The handover decision given to the networks with the highest rate and don’t consider the interaction between a required context and a provided one to make decision. III.
A. Structuring the decision problem into hierarchy model This step consists of building hierarchy model based on the context criteria. On the top level containing the decision goal and in intermediate level the context criteria according to their common characteristics. Fig. 1 gives an application of this model for network radio access selection. B. Establish local priorities of each criterion by making pair-wise comparison matrix To make pair-wise comparison we need to judge the context criteria two by two and indicate how many times a criterion is more important than the other one. Each of these judgments is assigned a scale number as shown in Table I. The intensity of importance use 9 point scales to convert these judgments to numerical priorities for every context criteria. For example if context A is strongly important than context B and we assign it the scale 5, then context B must be less important than context A and we assign it 1/5. The matrix is inversed with respect to the main diagonal that is equal to 1 because the diagonal represent the same context criteria compared with it. The local priorities are established by calculating the principal eigenvector of the pair-wise comparison matrix as given by AW= maxW, with A is comparison matrix, max is the largest eigenvalue of A and W is the corresponding eigenvector. The eigenvector are then normalized to become the priorities vector. C. Measurement of consistency of the comparisons and the overall weight The Consistency Ratio (CR) is calculated to validate comparative judgment phase at each pair-wise comparison matrix and to ensure reliability in determining the weight of a set of context criteria. It is calculated by (1) and (2)
CONTEXT AWARENESS
The context-awareness is all type of information presented in the environment which describes the situation of the user required service and the network in terms of location, time, user activity and network provider [12-13]. It can be got back from a big variety of sources and may be formed by various metrics such as network availability, network capability, bandwidth, signal strength, user preference, access cost, quality of service, security, mobility location, user activity, etc. Every application running on the MS has its own QoS requirements that can influence the handover decision regardless of network status. By allowing the user to choose a preferred network, the system is able to take into account a specific requirement of each application. The demand for context-awareness in heterogeneous radio access networks is related to the QoS specification. The context criteria that can be considered in our work: Received signal strength (RSS), Bandwidth, Data Rate, Round Trip Time, Latency, Jitter, Reliability, Cost and Security. IV.
ܴܥൌ
ூ ோூ
ܫܥൌ ሺ௫ െ ݊ሻȀሺ݊ െ ͳሻ
(1) (2)
when CI is the Consistency Index, max is the largest eigenvalue of A, n is the matrix size and RI Random Consistency Index. Saaty in [16] has provided average consistencies (RI values) of randomly generated matrices for a simple size of 500. The value of Consistency Ratio is acceptable if it’s equal or less than 0.10 for n5. Use the priorities obtained from the comparisons matrix to affect weight to all context criteria. Multiplying the local priorities of the sub-criteria by their parent’s criteria to obtained the global priorities.
ANALYTIC HIERARCHY PROCESS
The Analytic Hierarchy Process (AHP) is developed by Prof. Thomas L. Saaty [14-15]. It is popular and widely used, especially to resolve decision situation problem when multi-criteria are involved such as: choice, ranking, prioritization, resource allocation, quality management… The goal might be to establish a consistent way through pairwise comparison to allow users to judge the relative weight of each criteria (or sub-criteria). The procedure for using the AHP consists of the following steps as described in [14-15]:
Figure 1. AHP hierarchy model for network selection
494
TABLE I.
FUNDAMENTAL PAIR-WISE COMPARISON SCALE FOR AHP [16]
Intensity of importance 1
Definition Equal importance
9
Weak importance of Ai over Aj Essential or strong importance Demonstrated importance Absolute importance
2, 4, 6, 8
Intermediate
3 5 7
V.
criterion by their weight as follows: positive power for benefit criterion and negative power for cost criterion. Instead of addition, WPM use multiplication to assign NSFi for each network such as:
Description Element Ai and Aj are equally important Experience and Judgments slightly favour Ai over Aj Experience and Judgments strongly favour Ai over Aj Ai is very strongly favoured over Aj The evidence favouring Ai over Aj is of the highest possible order of affirmation When compromise is needed, values between two adjacent judgments are used
ܰܵܨ ൌ ςୀଵ ܽ ௐ VI.
MULTIPLE ATTRIBUTE DECISION MAKING
ൌ
ݓ
ǥ
ݓ ሿ
ೕ ିೕ
(6)
ଵଵ ڮଵ ۍ ڭ ڰ ڭ ൌ ێ ǥ ଵ ێ ۏଵ ǥ
ଵ ې ڭ ۑ (7) ܰܯܥ ۑ ے At this stage, our normalized matrix CMNPoAi can differentiate three types of numerical values for context criteria as shown in Fig. 2. • Positive value: where context criteria are widely satisfied (aij > ୨ ୲୦ ). • Zero: where context criteria are satisfied (aij = ୨ ୲୦ ). • Negative value: where context criteria are unsatisfied (aij < ୨ ୲୦ ). The goal of this step is to ignore the network that not satisfies some required context criteria of mobile user. These criteria are chosen based on service classes under running. For example connection is required if the RSSI and bandwidth are higher than the threshold. Then, remove the networks which not meet these contextual criteria. After filtering, we shall have an optimal satisfied solution but our objective is to choose the best solution. Thus it’s time to introduce the weight factor to select the PoA providing the highest performance of the context criteria. We suppose that the obtained rate for the rest of networks is given by (8)
(3)
The SAW model evaluate all networks and make decision by the use of (4) to ranking all candidates networks. The performed PoA is the highest value of Network Selection Function NSFi. ܰܵܨ ൌ σୀଵ ܹ
ೕ ିೕ
Where aij denotes context criteria j of candidate network i, ୨ ୲୦ is a threshold value or acceptable value for the jth criteria for all networks and ୨ is the highest value for benefit criteria (the lowest value for the cost criteria). The CMNPoAi matrix is given after normalization by (7).
A. Simple Additive Weighting (SAW) decision model SAW [6] is the simplest MADM method for ranking all networks detected. To make a decision, we need to follow the following steps: Construct the context matrix CMPoAi of networks for all context criteria. We have two types of context criteria, benefit and cost criteria. For the benefit criterion, the best value is the largest value, it have the highest acceptable value such as: RSS and bandwidth. For the cost criterion, the best value is the lowest value, it have the lowest acceptable value such as: delay and reliability. Thus, the context matrix CMPoAi is normalized and multiplied by their weight vector. The weights vector for every class of service is given by AHP approach. Where Wj denotes the weight of criteria j, and their sum is equal to 1 as in (3): ǥ
THE PROPOSED METHOD OF RADIO ACCESS NETWORK SELECTION
A. First Step: Ranking Scheme Ranking network with WPM or SAW method is computationally expensive especially in add or remove of networks. In order to reduce unnecessary computation, we propose to eliminate networks which do not satisfy a minimum user requirement. This purpose based on introducing a threshold value of the required context in the networks ranking stage. The steps are described as follows: After construction of the context matrix CMPoAi of networks with all context criteria, normalization is performed by the use of (6).
In heterogeneous environment the selection of suitable target network is the major problem when taking several criteria. The MADM is the best known method to resolve this problem. It evaluates the detected Point of Attachments (PoAs) in terms of multi-criteria. However, a variety of model is proposed in the literature to ensure successfully the use of different context-aware in decision making.
ܹ݄݁݅݃ݐ௦௦ ൌ ሾݓଵ
(5)
(4)
B. Weight Product Method (WPM) decision model The WPM [6] it’s another method to make decision where several criteria is involved. It’s similar to SAW model. The main difference is in the treatment of the benefit and cost criteria and instead of additive operator when ranking the networks there is multiplicative one. The context matrix CMWPMi is given with raise the power of each
ܰܵܨ ൌ σୀଵ ܹ
495
(8)
VII. NUMERICAL EXAMPLES AND EVALUATION The demand to perform handover in heterogeneous wireless networks is related to the context-aware. In this study, we consider a mobility scenario, composed by seven Point of Attachment uniformly distributed on a highway. We will assume that the MN running VoIP application is in position A served by PoA0 and this move to the position B. The idea of this purpose is to evaluate handover decision making in a scenario where the context metrics from the serving sector continuously decreases whereas the context metrics from the target sector continuously increases. The MN will thus pursue its connection with PoA0, by arriving at the point B, the MN detects the presence of six candidate networks PoA1, PoA2, PoA3, PoA4, PoA5 and PoA6. The handover is initiated by notifying the currently PoA0. This last one contacts the target PoA and informs it about the request of the MN and start exchanging messages. Based on the preference sent by the mobile node, the context provided by the target PoA is shown in table II (The measures of the attribute values for the candidate networks randomly vary according to the range of each context criteria). Various methods are executed to ranking the candidate networks. The MN running VOIP application, the vector of weights priorities for each context criteria can be given by AHP method. The numerical application of SAW and WPM methods defined in the previous section gives results in table III. Our solution is to introducing the threshold value of the required context in the ranking of networks. After normalization of the context matrix CMNPoAi as shown in Fig. 6, ignore networks that not satisfy some context criteria of mobile user. For example connection is required if the RSSI and bandwidth are higher than the threshold value. We can see here that the point of attachment number 3 and 4 are negative values for the critical criteria (RSS and Bandwidth), so it is below the threshold values. Then, remove of this networks which not meet these contextual criteria and reduce
Figure 2. Normalized Step
B. Second Step: Decision Scheme The second part of our proposal is to guarantee the QoS perceived by a given user. After ranking the rest of networks, the sensitive criteria, which can represent the impact of performance of each service class, have to be reviewed. As illustrated in Fig. 3, the Mobile Node (MN) initiates handover procedure due to the degradation of the quality of connection or to improve the quality by other networks with higher performance. The mobile node sends its preference to all detected networks. Each candidate networks computes its NSF and sends it back to mobile node. In the related work, perform handover to the highest NSF is triggered, but in our purpose the target before switching to the chosen network is to verify for the sensitive criteria that can impact the quality of service perceived in the end users. If the sensitive criteria of the selected network are upper than threshold value, then this selection should be held. Else, check the performance of network with NSFi-1 ranked after. Finally, another test should be handled in order to verify if the quality offered by the selected network is higher than the current one, since the handover is launched for quality improvement.
TABLE II. Criteria RSS Bandwidth Data Rate RTT Latency Jitter Reliability Cost Security
PoA1 1 0.4054 0.1193 0.452 0.6272 0.4054 0.418 1 0.3543
TABLE III. Candidate Networks PoA1 PoA2 PoA3 PoA4 PoA5 PoA6
Figure 3. Handover Decision Scheme
496
THE CONTEXT PROVIDER
PoA2 0.6281 0.4054 0.2991 1 0.6272 0.1643 0.418 0.3816 0.3543
PoA3 0.4126 0.1643 0.2991 0.2779 0.1476 0.4054 1 0.1455 1
PoA4 0.1298 0.1643 1 0.2779 0.382 0.4054 1 0.3816 0.1881
PoA5 1 1 0.2991 0.2779 1 1 0.125 0.3816 0.1881
PoA6 0.6281 0.4054 0.1193 0.452 0.2324 0.1643 0.125 0.1455 0.3543
APPLYING SAW AND WPM MODEL SAW NSFi
WPM NSFi
Ranking
0.594866 0.456294 0.318033 0.389894 0.746898 0.551031
0.514405 0.406423 0.263916 0.318175 0.631492 0.413352
2 4 6 5 1 3
unnecessary computation. The ranking value of the rest of candidate networks is very close to the other methods as shown in table IV. After ranking the candidate network and by comparison of results as depicted in Fig. 4, we notice that our proposal has the same behavior (ranking) as other methods. The benefit is avoiding the processing delay caused by unnecessary computation of networks which do not ensure connection. Before switching to the chosen networks, there is necessity to test the selected network in order to avoid multiple handover (Ping Pong effect). Briefly, low network performance can affects the user perceived quality but good performance quality does not imply high satisfaction to the end users. So with context awareness and each class of service features, we can define the sensitivity context criteria that have an important impact on the user’s satisfaction. As shown in Fig. 7, we can see that the variation of the weighting has a direct impact on the network selection specially the ranking of detected networks. Based on this variation we can define the sensitive context criteria. TABLE IV. Candidate Networks PoA1 PoA2 PoA3 PoA4 PoA5 PoA6
We compare our proposed algorithm Handover decision scheme with SAW method. Fig. 5 show that with SAW we note a ping pong effect (multiple handover decision) between PoA 1 and 2 which can block the communication and degrade the perceived QoS. This event is caused by the bad selection of the network with the highest NSF which does not satisfy the end users and then initiate another handover. But with our proposed scheme we note that the mobile node is attached to PoA2 avoiding the ping pong effect due the test which check for the sensitive criteria ensuring user satisfaction. Our algorithm can avoid unnecessary handover by testing sensitive context criteria that impact application constraint under running in the end users. VIII. CONCLUSION A VHO mechanism is a procedure of selecting the best access Point of Attachment among different RATs providing better link quality than the current one. In this paper we proposed a modification of the Simple Additive Weighting called Enhanced-Simple Additive Weighting (E-SAW) for ranking networks. Our contribution consists of avoiding the processing delay caused by unnecessary computation over available radio access technologies which do not ensure connection performance. So with such pre-selection, we avoid multiple handover decision generating a ping pong effect caused by the inefficient network choice according to end user required transmission performance. In future work, we enhance our proposed idea by enlarging our study handover execution.
APPLYING E-SAW MODEL NSFi
Ranking
0.285239 0.043622 Eliminated Eliminated 0.608276 0.283816
2 3
1 4
SAW Proposed Algorithm
CSAW CWPM
0,8
6
CE-SAW
0,7
5
Number of PoA
0,6
Ranking
0,5 0,4 0,3
4
3
2
0,2 0,1
1
0,0 PoA1
PoA2
PoA3
PoA4
PoA5
0
PoA6
20
30
40
Figure 5. Network selection for various class of service
Figure 4. Comparative Results
CMNPOAi =
10
Time (s)
PoAs
RSS
Bandwidth
D-Rate
RTT
Latency
Jitter
Reliability
Costs
Security
1 0.3669 0 -0.4814 1 0.3668
0 0 -0.4054 -0.4055 1 0
-0.2565 0 0 1 0 -0.2565
0.2411 1 0 0 0 0
0 0 -1.2864 -0.6577 1 1
0 -0.4054 0 0 1 1
0 0 1 1 -0.5034 -0.5034
1 0 -0.3818 0 0 -0.3817
0 0 1 -0.2574 -0.2574 -0.2574
Figure 6. Context Matrix CMNPoAi with E-SAW
497
Legend X axis: Criteria Y axis: Weight
Figure 7. Performance Sensitivity for context criteria: Goal Network Selection
REFERENCES [1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
M. Kassar, B. Kervella and G. Pujolle, “An overview of vertical handover decision strategies in heterogeneous wireless networks”, Journal Computer Communications Volume 31 Issue 10 pp. 26072620, June 2008. H. Y. Huang, C. Y. Wang, and R. H. Hwang, “Context-Awareness Handoff Planning in Heterogeneous Wireless Networks”, Lecture Notes in Computer Science, Volume 6406 pp. 430-444, SpringerVerlag Berlin Heidelberg 2010. P. TalebiFard and V. C.M Leung, “A Dynamic Context-Aware Access Network Selection for Handover in Heterogeneous Network Environments”, IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS) pp. 385-390, June 2011. J. M. Barja, C. T. Calafate, J. C. Cano and P. Manzoni, “An overview of vertical handover techniques: Algorithms, protocols and tools”, Computer Communications, Vol 34, No. 8 pp. 985-997, June 2011. X. Yan, Y. A. Sekercioglu and S. Narayanan, “A survey of vertical handover decision algorithms in Fourth Generation heterogeneous wireless networks”, Journal Computer Networks: The International Journal of Computer and Telecommunications Networking Volume 54 Issue 11 pp. 1848-1863, August 2010. K. Savitha and Dr. C. Chandrasekar, “Vertical Handover decision schemes using SAW and WPM for Network selection in Heterogeneous Wireless Networks”, Global Journal of Computer Science and Technology Volume 11 Issue 9 Version 1.0 pp. 19-24, May 2011. Hongwei Liao, Ling Tie and Zhao Du, A Vertical Handover Decision Algorithm Based on Fuzzy Control Theory, Proceedings of the First International Multi-Symposiums on Computer and Computational Sciences (IMSCCS'06) pp. 309-313, June 2006. Rashid A. Saeed, Hafizal Mohamad, Borhanuddin Mohd. Ali and Mazlan Abbas, WiFi/WiMAX Heterogeneous Seamless Handover,
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
498
Third International Conference on Broadband Communications, Information Technology & Biomedical Applications, pp. 169-174, December 2008. Z. Dai, R. Fracchia, J. Gosteau, P. Pellati and G. Vivier, Vertical handover criteria and algorithm in IEEE 802.11 and 802.16 hybrid networks, IEEE International Conference on Communications 2008 (ICC '08), pp. 2480–2484, Mai 2008. Xu Haibo, Tian Hui and Zhang Ping, A novel terminal-controlled handover scheme in heterogeneous wireless networks, Journal Computers and Electrical Engineering, Volume 36 Issue 2 pp. 269– 279, March 2010. K. Piamrat, A. Ksentini, J. M. Bonnin and C. Viho, “Radio resource management in emerging heterogeneous wireless networks”, Computer Communications, February 2010. Ricardo Matos and Susana Sargento, Context-Aware Connectivity and Mobility in Wireless Mesh Networks, Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, Volume 32 Part 2 pp- 49-56, 2010. Thomas Hofer, Wieland Schwinger, Mario Pichler, Gerhard Leonhartsberger, Josef Altmann and Werner Retschitzegger, ContextAwareness on Mobile Devices - the Hydrogen Approach, Proceedings of the 36th Annual Hawaii International Conference on System Sciences (HICSS'03), Track 9 Volume 9, 2003. Thomas L. Saaty, Decision making with the analytic hierarchy process, International Journal of Services Sciences (IJSSCI), Vol. 1, No. 1, pp. 88-98, 2008. Thomas L. Saaty, How to make a decision: The Analytic Hierarchy Process, European Journal of Operational Research 48, pp. 9-26, North-Holland 1990. R.Ramanathan, A note on the use of the analytic hierarchy process for environmental impact, Journal of Environmental Management 63, pp. 27-35, 2001.16
