CHAPTER 2: LITERATURE REVIEW. Wireless network is a dynamic research domain. Already considerable work has been

Transcription

1 13 CHAPTER 2: LITERATURE REVIEW Wireless network is a dynamic research domain. Already considerable work has been carried out to improve the sagging performance of IEEE WLAN in the areas of quality of service architecture, admission control, voice and video streaming, scheduling, bandwidth allocation and load balancing [3]. This thesis is concerned with the literature from the following research domains. QoS Architecture Admission Control Scheduling Bandwidth Allocation Load Balancing After analyzing the literature, it would be clear that legacy IEEE WLAN is not appropriate for delivering real-time bandwidth greedy applications. The QoS enabled IEEE e standard has no description as to how to adjust the four medium access parameters viz; AIFSN, CWmin, CWmax and TXOP associated with each access categories. The proper tuning of the parameters of these four access categories will definitely improve the performance of the wireless medium. Several algorithms proposed by other researchers have attempted to extend the provision of guaranteed services are deeply analyzed in this chapter. The enhancements proposed in this thesis aims to deliver a quality of service improvement by ensuring the realization of the quality of service aspects.

2 QoS ARCHITECTURE In wired and wireless communication systems, the quality of service becomes an essential requirement. This is because the information access becomes more and more ubiquitous. Many researchers have proposed robust solutions to address this end-toend QoS assurance in IEEE WLAN. By an extensive literature survey, it is clear that in the currently available end-to-end QoS architectures, some important characteristics like cross layer integration, end-to-end integration, re-configurability and modularity are still not adequately addressed. M. Li, H. Zhu, I. Chalmtae and B. Prabhakaran, [36], in their research paper End-to- End QoS Framework for Heterogeneous Wired Cum Wireless Networks proposed WLAN flow reservation protocol (WRESV). This is an admission control and flow reservation control mechanism for IEEE WLAN for accessing signals from both wired and wireless network and offer to provide the required quality of service. The RSVP-WRESV combination technique considers the important characteristics of RSVP and the wireless channel. The cross layer interaction mechanism provided in the framework reduces the overhead of message mapping and provides adequate QoS provision for real-time flow in diverse wired or wireless networks. The framework is unable to address the requirements such as seamless end-to-end integration, reconfigurability and modularity. H. Nguyen, R. Rivas, and Nahrstedt, idsrt [26], integrated Dynamic Soft Real- Time Architecture for Critical Infrastructure and Delivery over WLAN. The authors proposed idsrt, it includes a processor scheduler (DSRT), a network scheduler (EDF-Earliest Deadline First), a distributor, and a coordinator at the middleware layer to integrate all schedulers. It integrates all these scheduler functions, to provide endto-end guarantee of packet transfer. This idsrt framework resides and functions in

3 15 the middleware, network and operating system layers. Network traffic types like realtime and best-effort service is measured on a node by DSRT. The network packet scheduling is performed by iedf. It takes implicit contention steps according to the EDF policy. The idsrt was designed to support fine-grained end-to-end minimum delay guarantees. W.He, and Nahrstedt, [27], in their research paper Impact of Upper Adaptation on End-to-End Delay Management in Wireless and Ad-hoc Networks, proposed a quality of service system that provides end-to-end bandwidth guarantee using the service differentiation approach. In the reference architecture it has a bandwidth monitor in the link layer to monitor the bandwidth requirement, a priority adaptor and a packet classifier in the middleware layer and a waiting time priority (WTP) packet scheduler at the network layer. The key operating principle of M-WTP is a mapping of packet level priority to the end-to-end semantics and feed-back control mechanism to adapt priority for end-to-end performance. It is also noted that there is a large jitter and slow adaptation. This results in a weak end-to-end delay guarantee ADMISSION CONTROL Jen-Jee Chen, Ling Lee, Yu-Chu Tsang [49], in their research paper Integrating SIP and IEEE e to Support Hand-off and Multi-Grade QoS for VoIP over WLAN Applications, the authors considered to provide QoS for the VoIP signal by multigrade QoS and hand-off and rate adaptation. The solution proposed includes call admission control and resource management by integrating Session Initiation Protocol (SIP) and QoS mechanism of IEEE e together. This process dynamically adjusts the resource distribution among existing calls by controlling their supporting Packatization Interval (PI) and codecs. Admission control method used in this paper not only accepts more calls, but also supports the existing calls under bad channel

4 16 conditions. This reduces the blocking rate for newly arriving calls and respectively decreases the call dropping rate for hand-off calls. The proposed mechanism is tested only for VoIP signals and not for all types of real-time traffic and BE services. Venkatakrishsan and Selakennedy, [51], An Enhanced HCF for IEEE e Wireless Networks, proposed an enhanced hybrid coordination function (e-hcf) to improve the performance of IEEE WLAN. From their comparative analysis about EDCA and HCF it is clear that the performance of EDCA is satisfactory only for the networks with a few numbers of stations. Hybrid coordination function performs well for a network with large number of stations. They also have recognized that bursty traffic and variable bit rate signals received considerably low QoS from both EDCA and HCF and this leads to starving in large and busy networks. The proposed e-hcf uses class based queuing (CBQ) techniques. This provides fairness in the bandwidth allocation [3]. It exhausts the hybrid coordinator (access point) resource and prevents a specific traffic category (like voice) from obtaining all the transmission opportunities (TXOPs). To minimize the starvation and delay in medium access, they have proposed a percentage based ratio to the admission control mechanism. This assigns a portion of the total bandwidth to every traffic category and allocates the hybrid coordinator resource to them. It reserves a certain portion of bandwidth for a specific class of traffic. This mechanism enhances the throughput of the WLAN and hybrid coordinator s resource allocation. Still the bandwidth may not be optimally utilized to provide the required quality of service. Gao, D. Cai J and Zhang L, [19], Physical Rate Based Admission Control for HCCA in IEEE e WLANs, suggested an admission control mechanism, known as a physical rate based admission control scheme (PRBAC) to enhance the performance of IEEE WLAN by considering important features of wireless medium and the

5 17 mobility of the wireless nodes. The key objective of any admission control system is, to schedule the traffic into a specific class of service to provide a better quality of service of all admitted flow and at the same time utilizing the wireless medium and the resources optimally. The PRBAC technique considers the variation of physical channel rate which is caused by the mobility of the wireless node and characteristics of the medium. This is achieved by an admission control mechanism to provide long term average physical rate, and at the same time the required TXOPs to discrete quality of service stations, based on their instantaneous physical rates. Using the proposed PRBAC scheme the authors claim that it allows more traffic streams to access the wireless medium than the standards scheduler, and increases the TXOPs. This technique uses the average physical rate of the channel for admitting the traffic, also select some quality of service sessions and drop the packets arbitrarily. However the process of arbitrarily removing TXOP for a quality of service station may degrade the performance of a WLAN. Didi, et.al, [12], have suggested an admission control mechanism. It uses the number of BE traffic from a quality of service stations and their position in a WLAN as the controlling parameters. To provide channel access for the best-effort traffic, and prevent the low priority signals from starvation, this algorithm increases the CW, CW and AIFSN only when the throughput reaches 70% of the maximum value. The max TXOP is decided based on the current transmission rate and the position of the quality of service stations, rather than considering the minimum transmission rate. The hybrid coordinator re-calculates the TXOP provided to the quality of service stations with VBR load, when the packet drop reaches a threshold. Correspondingly if the collision rate exceeds a particular limit (threshold value), then the number of quality of service stations transmitting the frames are minimized to retain the quality of service for the min

6 18 existing audio and video traffic streams. The proposed mechanism provides better performance in terms of data rate, transmission delay and collision compared to HCF controlled channel access. However if the QSTA are moving, the quality of service access point needs to calculate the current position and the quality of service required by the wireless stations regularly, this increases the overhead significantly. Xiao Y., Li F. H., Li M., Jingyuan Z., Li B., and Hu F, [55], Dynamic Budget Partition Scheme for Integrated Voice/Video/Data Traffic in the IEEE802.11e WLANs, proposed the admission control based bandwidth partitioning scheme for multimode traffic over IEEE e WLAN. The available bandwidth is allotted to access the medium by static and dynamic bandwidth allocation mechanism among the various traffic categories based on the type of the traffic. The simulation and performance analysis used indicates quality of service as average throughput for multimedia traffic [48], total throughput, number of accepted and active flows, etc. Simulation results show that the bandwidth allocation [3] by dynamic scheme is better than the static scheme. It is conducted for CBR traffic and not for the real life VBR traffic. The enhanced quality of service schemes of IEEE802.11e EDCA MAC mechanism provides prioritized channel access for real-time applications. It does not describe the mechanism to ensure strict QoS required by the real-time traffic. A number of techniques with different QoS enhancement schemes such as, admission control, rate control, proper tuning of IEEE e parameters (AIFSN, CW, min CWmax and TXOP) etc. have been proposed to enhance network performance and thus to achieve accepted quality of service. G. Kularni, A.Nandan, Mario Gerla and Mani Srivastava [29], A Rate Adaptive MAC Protocol for MIMO Based Wireless Network, proposed a rate adaptive MAC protocol for WLAN with MIMO links. The authors proposed a technique to maximize

7 19 the achievable data rate by minimizing the BER, using a feedback mechanism for the transmitter to obtain the rate selection setting from the receiver. The proposed scheme is not tested for different channel models. Duk Kyung Kim Inha, David Griffith, and Nada Golmie, [15] in their research A Call Admission Control Scheme for Heterogeneous Wireless Network, the admission control mechanism and bandwidth reservation issues are largely been investigated in many studies. Authors in this research paper proposed to adopt minimum bandwidth requirements as an important criterion for admission control for all classes of services. This mechanism can provide more connections and large number of users contending to access the AP in a MIMO system. These users are allowed to access the MIMO channel, without considering the priorities of the real-time traffic. This may cause a relatively low QoS performance. Dusit Niyato et, all, [16], in the Joint Admission Control and Antenna Assignment for Multiclass QoS in Spatial Multiplexing MIMO Wireless Networks, the authors detailed about the objectives of the admission control algorithm to enhance the performance of WLAN, for different traffic categories. The proposed technique is based on the constraint connection blocking probability and average per-connection throughput. Constraint Markov decision process model is proposed to obtain the optimal decision on antenna assignment and admission control. Radio resource management with adaptive modulation technique is considered for better channel utilization. The proposed scheme outperform for the best-effort service. If the traffic categories include real-time traffic the performance decreases.

8 SCHEDULING Bin Muhamad Noh, SuzukandTasaka, [59], Guaranteeing QoE in Audio-Video Transmission by IEEE e HCCA. In this paper they proposed two scheduling techniques (the static and dynamic scheduling). The static scheduling scheme is more suitable for the traffic with CBR, and it grants the TXOPs with better channel access than the standard schedulers. TXOP in static scheduling is obtained by multiplying the mean data rate of a specific traffic, and the parameter α, which is a variable and its value is greater than 0. The dynamic scheduling scheme is also known as multimedia priority dynamic scheduling (MPDS). This technique allocates TXOPs for the traffic streams of VBR traffic. The channel access is provided based on the queue length of quality of service stations obtained in the previous service interval. The authors compare the performance of both static and dynamic scheduling scheme, with respect to the user level and application level quality of service. The simulation result shows that a high user level QoS can be obtained by using static scheduling. The MPDS produces a better performance than the static scheduling and standard scheduler only if the number of stations is less. Lagkas, et.al, [31], Priority Oriented Adaptive Control with QoS Guarantee for WLANs, suggested priority oriented adaptive control mechanism with QoS guarantee (POAC-QG) to schedule the real-time and background traffic period. The POAC-QG functions in contention based access mechanisms such as EDCA [19]. The POAG-QG uses TDMA based channel access method. The real-time duration is divided into time slots to keep QoS stations synchronized. For maintaining the synchronization, the QSTA will get updated information like time interval, source and destination information for the packet transmissions. The hybrid coordinator updates

9 21 this information to the QSTAs using beacon frames. The admission control algorithm considers high priority traffic [48], if the throughput is sufficient to serve with maximum quality level. If the available bandwidth is unable to accommodate and serve with maximum quality level, then low priority traffic streams are served [53]. Lastly if the traffic stream cannot be accommodated with minimum quality level required, it is disallowed and the succeeding priority flow is verified. A part of the previously processed higher priority traffic is lowered when the available bandwidth is unable to accommodate traffic streams with minimum priority level [48]. This helps to provide a portion of the bandwidth for newly arriving flow. The analysis of the result shows that the POAC-QG scheme enhances the performance of WLAN with improved delay and less jitter for streaming voice and video traffic compared to HCCA. POAC-QG is a different technique for HCF controlled channel access and does not support traffics other than multimedia traffic. Allocating the bandwidth for high priority traffic streams will bring the low priority traffic in to starvation. Gozalvez.et.al, [9], An Efficient HCF Scheduling Mechanism in Mixed Scenarios, the algorithm Hybrid HCCA-EDCA Centralized Scheme is with an objective to enhance the QoS demand of WLAN for VBR traffic. In this, the bandwidth reserved for the up-link traffic is compensated for the requirements of downlink bandwidth and then the size of TXOPs are increased. A token based algorithm is used to serve the best-effort traffic. The QAP assigns tokens to QSTA based on their data transmission rate. The QoS access point transmits to the QoS station with the maximum number of tokens. The real-time traffic is allowed to access the channel compared to the besteffort traffic in the downlink scheduler. All uplink best-effort traffic is served in the next contention period through EDCA. It is not transmitted during contention free period. The simulation result shows that this algorithm can enhance the performance

10 22 of best-effort traffic without interfering with the service quality of real-time traffic. However the extra throughput assigned for variable bit rate traffic without having the details about the amount of throughput required for the QSTAs, wastes the channel bandwidth. Chen and Yeh, [56], An Adaptive Polling Scheme Supporting Audio-Video Streaming in WLANs, In this research the authors suggested a polling [56] based scheduler Adaptive Time stamp polling (ATSP) method, for improving the QoS and identifying the silent mode of a VoIP traffic to minimize the number of times a silent station is polled. The proposed algorithm calculates the exact time slot to poll the QSTA, which have data ready to transmit, and polls the data ATSP. The channel access delay is minimized for small interval called short interval polling (SIP). This polling is done at the beginning of their transmission duration. The algorithm also proposes a talk-spurt and silence detection algorithm to minimize the wastage of bandwidth. The response from the QoS station contains a quality of service data frame for the QoS CF-poll frame from the hybrid coordinator. The hybrid coordinator considers that the traffic stream is in the talk-spurt state, and it uses the packet interarrival time as the pooling arrival time. This period is considered for deciding the pooling interval. If the QoS station sends responses with quality of service-null frames, three or more times repeatedly, then the hybrid coordinator will consider that the traffic stream is in the silent state. Under silent state the polling time for the traffic stream is increased up to 300 milliseconds, to minimize the overhead. The performance analysis shows that using the proposed algorithm, there is an improvement in the QoS parameters like throughput, delay and jitter. However this mechanism increases the computational complexity of the scheduler. The hybrid coordinator needs to perform the calculation related to the polling period for every

11 23 traffic stream, whenever there is a change in the polling list [56]. This increases the overhead of the hybrid coordinator LOAD BALANCING Ioannis, Papanikos and Micheal Logothetis, [22], A Study on Dynamic Load Balance for IEEE b Wireless LAN, proposed load balancing procedures to function in two separate levels. The access points are either distributed across the channels or located in the same channel. This is indicated by considering the channel location and the RSSI value of the neighboring access point. The wireless stations are distributed across all the available access points based not only on the RSSI measurements associated to the access point but also other link quality measurements. The proposed algorithm functions in three distinct levels viz; in channel access and auto-selection level, the station decision level, and the link observation level. The implementation result of the algorithm shows a balanced distribution of the number of wireless stations to the access point and an enhancement in the overall network performance. However the algorithm needs an enhancement with the decision levels among the wireless stations, the resource availability in the access points and the frame error rate. Shanjun Chung.et.al, [45], A Distributed Constraint Optimization Algorithm for Dynamic Load Balancing in WLANs, proposed the technique for dynamic load balancing mechanism and to respond to the volatility in a wireless local area network. The authors proposed a load balancing framework for wireless local area network. This is a multi-agent system. It maps the load balancing problem in a wireless local area network into a distributed constraint optimization problem. The new dynamic load balancing mechanism is a distributed optimization algorithm called Dynamic Load Balancing Distributed Pseudo-tree Optimization Procedure (DCB-DPOP). The

12 24 algorithm mainly emphasizes on pseudo-tree repair instead of pseudo-tree reconstruction, each time when the load on the WLAN changes. The distributed optimization algorithm DLB-SDPOP has functionalities such as, self-stabilizing pseudo-tree repair mechanism, up-keeping of the effected wireless stations in the original pseudo-tree and repair instead of re-constructing the entire pseudo-tree; each time a perturbation needs to be handled. This minimizes the complexities in load balancing mechanisms. The simulation result shows that the proposed multi-agent approach optimally allocates the mobile stations under each access point. This algorithm efficiently handles up to 50% perturbation in real-time environment and outperforms DPOP. However the performance of the algorithm reduces in an ESS with large number of closely located access points. Hector Velayos.et.al, [50], in their Load Balancing in Overlapping Wireless Local Area Network Cells proposed the load balancing scheme for completely distributed architecture based on agents functioning in every AP. The agent updates the information about the load conditions in every access point, via the interconnected wired network and select a particular wireless station, that have to be moved to the neighboring AP by running a hands-off process to balance the load [15]. The entire process is considered as load balancing cycle (LBC). The LBC always starts collecting the balancing index and finishes balancing the current access point load information to the neighboring access points through Ethernet backbone network. The algorithm functions only if the load in the access points is not balanced. It calculates the average load and determines the state of the access points. Using this information the access rules are updated to balance the load. The algorithm was tested by implementing it on a Linux system. Result showed that, there is an enhancement in the performance with increased throughput and reduced delay. Only the balanced and

13 25 under-loaded access points accept the new stations. The overloaded access points run the hands-off process to move the connected wireless stations until the load is balanced. However number of hands-offs and re-association of stations to the access points, increases the overhead SUMMARY There have been number of QoS proposals based on different IEEE standards for different traffic categories. Most of the solutions proposed for the QoS enhancement of IEEE802.11protocol. From the studies mentioned earlier it can be concluded that IEEE standard provides only best-effort traffic without proper service differentiation. As a result of this the real-time applications will suffer from poor performance, and without proper QoS mechanisms. The IEEE e with, EDCA MAC scheme supports prioritized channel access for real-time traffic in order to realize an acceptable QoS. A number of QoS enhancement mechanisms are introduced for IEEE e standard. This includes the admission control, rate control and proper tuning of the IEEE e parameters (AIFSN, CWmin, CWmax, and TXOP). Some of the relevant published works are described in this section. It can be concluded after receiving these studies that, they agree in general that by tuning these network parameters performance can be enhanced. This conclusion guided for this research to develop the new QoS enhancement mechanisms such as, QoS Architecture, Adaptive Dynamic Admission Control Technique for IEEE e, Priority Based Admission Control Mechanism Adaptive Modulation and Coding for IEEE n, and Dynamic Load Balancing in WLAN (described in chapter 4, 5, 6, 7and 8).