A Tool for Simulating IEEE e Contention-based Access

Transcription

1 A Tool for Simulating IEEE e Contention-based Access Andreas Floros 1 and Theodore Karoubalis 2 1 Dept. of Informatics, Ionian University, Plateia Tsirigoti 7, Corfu, Greece floros@ionio.gr 2 ATMEL Hellas S.A., Stadiou str., Platani, Patras, Greece tkaroubalis@athens.atmel.com Abstract In this paper, a simulation tool is presented, suitable for evaluating the performance of the e contention-based access e is a very recent amendment to the IEEE standard, describing the MAC layer enhancements required for Quality of Service (QoS) provision. Aim of the developed simulator is to provide an accurate and efficient tool for evaluating significant parameters of the e protocol, mainly in terms of traffic differentiation for QoS support. To evaluate the accuracy and demonstrate the use of the simulation tool, a number of simulations are performed, following the basic e interoperability test procedures. I. INTRODUCTION During the recent years, the wireless local area networks (WLAN) technology [1] has met a widespread market adoption, which is significantly increased due to the reduction of the wireless equipment cost and the continuous ratification of enhanced protocols and mechanisms for higher transmission rates and advanced security protection. However, provided the market trends for real-time multimedia traffic delivery, one of the major weaknesses of the networks is the lack of Medium Access Control (MAC) mechanisms with efficient Quality of Service (QoS) features. Towards this aim, the e amendment [2] is focusing on enhancing the MAC layer with advanced mechanisms for QoS support. As it will be explained in detail later in this paper, this is achieved following two different approaches: traffic prioritization and centralized, controlled access. In both cases, these enhancements significantly extend the development and the final wireless applications capabilities. As the main challenge is to optimize these new capabilities through the development of efficient QoS schemes, the development of new simulation tools and testbeds for QoS over WLAN represents a very critical and significant task. In this work, a novel software platform is presented that allows the fast and accurate simulation and evaluation of the e mechanisms that are realized using contention-based access and traffic differentiation. The rest of the paper is organized as follows: Section II presents a short overview of the protocol. In Section III, the e QoS enhancements are presented, focusing on the contention-based features, while Section IV provides the general description of the proposed simulation platform. Section V presents a test methodology for verifying the accuracy of the simulation tool and also includes a case study with the use of the platform. Finally, Section VI concludes and presents the future plans with respect to this work. II. IEEE CHANNEL ACCESS MECHANISM It is well-known that the IEEE specification defines the Distribution Coordination Function (DCF) as the mandatory wireless medium access mechanism. DCF represents a listen-before-talk mechanism, where each station (STA) is considered as an independent backoff entity that senses the medium during a DCF interframe space (DIFS). If the medium is idle, the STA additionally waits for a random time B. The backoff value B is randomly chosen in the interval [0, CW), where CW is the Contention Window with values in the range CW min CW CW max. CW min and CW max are constants defined by the protocol. If B=0, a packet transmission is initiated. Upon a successful transmission, a new backoff value is selected and the contention window is set to CW min, otherwise the CW value is doubled up to the CW max value. On the other hand, if the medium gets busy while a STA is decreasing the backoff timer, the backoff procedure is paused and is resumed after the medium is sensed to be idle for DIFS. Since more than one STAs may concurrently gain access to the medium, collisions may occur. Upon a collision detection obtained by an acknowledgment scheme, the CW value is doubled up to CW max and a new backoff value is selected. Upon traffic congestion or heavy link quality degradation, the collision rate is increased, resulting into significant throughput and delay degradation. In order to overcome these conditions, the specification also defines the Point Coordination Function (PCF). Under PCF, the transmissions are centrally coordinated using a polling scheme by the Point Coordinator (PC), located into the Access Point (AP), which has higher access priority, as it detects the medium status for PCF interframe space (PIFS), with PIFS<DIFS. Upon a poll reception, a STA may transmit a pending data packet. However, although the centralized nature of PCF introduces basic QoS features, a number of problems it raises, such as unpredictable beacon delays and unknown transmission durations [3], render it insufficient for practical QoS implementations. III. OVERVIEW OF E QOS ENHANCEMENTS In order to achieve true QoS performance, the e specification, apart from the legacy DCF and PCF, also defines an additional access coordination scheme termed as Hybrid Coordination Function (HCF), which consists of: a) the Enhanced Distributed Channel Access (EDCA), for differentiated, distributed access to the wireless

2 medium and b) The Hybrid Controlled Channel Access (HCCA) for parameterized, controlled access. Both access methods result into Transmission Opportunities (TXOPs) used by the QoS stations (QSTAs) for data transmissions. As EDCA is only considered here, a description of the HCCA channel access is out of the scope of this work, but can be found in [4]. A. EDCA Overview EDCA contention-based access achieves traffic differentiation by mapping 8 user priorities to 4 Access Categories (ACs), namely AC_VO (Voice), AC_VI (Video), AC_BK (Background) and AC_BE (Best effort). Each AC can be considered as an independent backoff entity that contend for medium access, using a set of EDCA parameter values (i.e. arbitration interframe space - AIFS, contention window - CW, maximum transmission length TXOP Limit) shown in Table I, under the legacy DCF rules. TABLE I DEFAULT EDCA PARAMETER VALUES AC CW min CW max AIFSN TXOP Limit b a/g AC_BK AC_BE AC_VI ms 3.008ms AC_VO ms 1.504ms The EDCA parameters are defined and announced by the Hybrid Coordinator (HC). For AC differentiation, it is very important that the same values are used by all backoff entities that are mapped to a specific AC. An EDCA TXOP is granted to an AC, provided that the medium is detected to be idle for: AIFS[AC] = AIFSN[AC] x aslottime+sifs (1) and that it s backoff counter B[AC] is zero. aslottime and SIFS are time intervals defined by the protocol. The backoff counter value B[AC] is randomly chosen following the normal DCF rules described in Section II, using the modified CW min and CW max values shown in Table I. A collision between two or more ACs belonging to different QSTAs is treated as in legacy DCF. On the other hand, in the case that two or more ACs within the same QSTA gain access to the medium, an internal collision occurs and the TXOP is granted to the highest priority AC. Due to the statistical, distributed nature of the EDCA channel access, in case of excessive data load and traffic congestion, no QoS guarantees can be practically conformed. In order to overcome this problem, as defined in [2], the EDCA-based transmissions may be subject to certain access restrictions in the form of an admission control mechanism for providing guarantees on the amount of time an admitted traffic will access the wireless medium. These requirements are defined by the serviced traffic upon admission in a Traffic Specification (TSPEC) element. The EDCA admission control scheme employed is a very significant parameter that directly affects the overall system performance, especially in the case of heavy traffic load and/or wireless channel degradation. It is generally applied on a) the QoS AP (QAP) side, providing guarantees on the amount of time an admitted traffic will access the wireless medium and b) the QSTA side, for monitoring whether the AC transmissions have exceeded the admitted time. Although the minimum functionality of the EDCA admission control is analytically defined in [2], the employment of advanced, adaptive admission control schemes can preserve the allocated EDCA access guarantees, even under unpredictable, time varying channel conditions. The description of an enhanced, effective admission control scheme that overcomes instantaneous channel degradations can be found in [5]. B. TXOP Continuation In order to improve the overall EDCA bandwidth management, the e specification defines an optional TXOP continuation scheme. Under TXOP continuation, a TXOP winner AC retains the right to access the wireless medium and may transmit additional packets consecutively. In any case, TXOP continuation is allowed, provided that the following four rules are satisfied: a) The previous packet transmission was successful. b) There is at least one packet available in the AC transmission queue. c) The duration of the next packet to be transmitted (including all overheads) must conform to the admission control limitations and rules. d) The total length of all consecutive transmissions allowed does not exceed the TXOP Limit value defined in Table I. In this Table, a zero value indicates that TXOP continuation is not permitted. If all the above conditions are satisfied, then a QSTA may commence transmission for a specific AC at SIFS after the completion of the immediately preceding frame exchange sequence. It must be noted that TXOP continuation is only permitted for the transmission of a packet belonging to the same AC, as the AC that was granted the initial TXOP. IV. SIMULATION PLATFORM DESCRIPTION Based on the previous description of the EDCA function, it is clear that the choice of the admission control type significantly affects the EDCA QoS performance. This performance can be further improved using the TXOP continuation mechanism. Hence, an important issue is the development of a tool that will help wireless product developers to design effective EDCA admission control and TXOP continuation schemes. This was the primary motivation for developing the EDCA simulation tool proposed in this work. The wireless EDCA simulator is a MS Windows-based application that simulates and evaluates the performance of typical wireless, contention-based channel access scenarios, including scheduling schemes and admission control strategies, based on the IEEE802.11e draft v.13.0 standard [2]. From the very early design stage, the following basic requirements were considered: user friendly, windows - based interface that will allow the fast definition of simulation scenarios, without requiring the knowledge of special description languages (such as TCL/TK). hybrid event and time driven simulation core for handling all protocol signaling and for improved simulation execution times.

3 support of all mandatory EDCA procedures including MAC-layer introduced transmission overheads. traffic source modeling based on TSPEC parameters (including burst modeling) defined by the Wireless Fidelity Alliance (WFA) in [6]. parameterized wireless channel error modeling. graphical interface for plotting measured averaged and instant throughput. performance evaluation using legacy and advanced QoS-related networking criteria. In order to perform a simulation, a scenario must be defined, using the graphical user interface shown in Fig. 1. A typical simulation scenario consists of: a) a number of activated QSTAs and their traffic streams (TSs) requesting service from the QAP. Currently up to 10 QSTAs and 4 TSs per QSTA are supported. A QSTA is activated when at least a TS is defined using the window shown in Fig. 2. As mentioned previously, the default TSPEC values are obtained from [6], however, the user is allowed to alter any of these default values. In the same window, the TS buffer length can be also set and the traffic source burst model can be activated. Fig. 3. QAP properties window After executing a simulation scenario, a detailed analysis can be reviewed (see Fig. 4) which contains information about the TXOPs granted (including all beacon transmissions), such as whether an external (or internal) collision has occurred and the updated relative backoff values, the TXOP starting time and length, the portion of TXOP employed for transmissions and the amount of data queued in the corresponding TS buffer. The instantaneous aggregated bandwidth value admitted for service is also appeared after every successful TS admission. All the above mentioned information can be exported in html format for easy future reference. Fig. 1. The wireless EDCA simulator application window Fig. 2. Traffic Stream properties window b) a QAP device. The QAP properties, including the EDCA parameter values described in Section III, the admission control type, as well as the QAP radio type can be defined using the window shown in Fig. 3. Apart from the admission control selection, the user can determine the total or per AC admission control limits (that is the % portion of the total bandwidth offered for admitted services only). c) various simulation parameters, such as the total duration and the channel error model in terms of packet loss probability. Fig. 4. Simulation analysis window Moreover, a number of metrics are calculated in per AC, TS and QSTA basis (see Fig. 5). Apart from the wellknown throughput and packet delay measurements, these include the amount of data overflowed, the number of TXOPs granted, the total collisions occurred and the service interval violations observed. The last metric represents a valuable QoS-related criterion representing the ability of EDCA to guarantee data transmissions, before the TS service interval defined in the TSPEC element expires. Fig. 5. Simulation statistics window

4 V. RESULTS A. EDCA traffic differentiation and fairness assessment In order to verify the wireless EDCA simulator accuracy, a sequence of tests was simulated, which are defined by the WFA for determining basic interoperability between different vendor products and for providing appropriate certification to commercial wireless equipment. The detailed test description is provided in [7]. Generally, the aim of the tests is to ensure that a) traffic differentiation is achieved among different ACs and b) access fairness between equal priority ACs is offered under any networking conditions. For the traffic differentiation test, two TSs belonging to different ACs and QSTAs were defined, as shown in Table II. According to the test definition, the high priority traffic does not exceed the link capacity, while the low priority stream provides enough traffic load, saturating the wireless link. The physical (PHY) transmission rate was set to 1Mbps for all cases, while two wireless link cases were considered: a) ideal wireless channel and b) nonideal wireless channel with 30% packet loss probability. Table II also shows the results obtained using the wireless EDCA simulator for both link quality cases. The ThroughputRatio parameter appeared in this Table is defined as the (%) ratio of the achieved throughput to the corresponding mean data rate. From these results, the following conclusions can be drawn: 1. Due to the channel saturation, the total number of TXOPs obtained by the QSTAs is almost constant in both wireless link cases. 2. In both cases, the high priority stream gets the bandwidth it needs to achieve is intended load, while the low priority traffic gets only the bandwidth left. 3. The high priority traffic packet delay is very low, even in the case of non-ideal wireless channel. Hence, taking into account the corresponding throughput performance, it is clear that the wireless EDCA simulator achieves the required traffic differentiation. TABLE II TRAFFIC DIFFERENTIATION TEST RESULTS Traffic G.711 CD-quality Voice Audio Mean/Total User priority 6 (VO) 2 (BK) - Mapping AC AC_VO AC_BK - MeanDataRate 166kbps 800kbps 966kbps Ideal wireless channel TXOPs granted Throughput 165.9kbps 669.4kbps 835kbps ThroughputRatio 99.99% 83.68% 86.48% Packet delay 0.158ms ms ms Non-ideal wireless channel TXOPs granted Throughput 165.9kbps 449.5kbps 616kbps ThroughputRatio 99.99% 56.19% 63.72% Packet delay 1.265ms ms ms For the AC fairness test, two identical traffic flows mapped to equal AC priorities were defined, as it is shown in Table III. The total intended load exceeded the channel capabilities, while the simulations were again performed for ideal and non-ideal wireless channel (with 30% packet loss probability). From these results it is clear that the wireless EDCA simulator additionally demonstrates the desired AC fair handling, as both streams get half of the link effective bandwidth. The same trends are observed in the non-ideal wireless channel case, although the throughput and packet delay performance is significantly degraded. TABLE III AC FAIRNESS TEST RESULTS Traffic CD-quality CD-quality Audio Audio Mean/Total User priority 5 (VI) 5 (VI) - Mapping AC AC_VI AC_VI - MeanDataRate 500kbps 500kbps 1Mbps Ideal wireless channel Throughput 420.1kbps 417.9kbps 838kbps ThroughputRatio 84.02% 83.58% 83.80% Packet delay ms ms ms Non-ideal wireless channel Throughput 308.5kbps 310.0kbps 619kbps ThroughputRatio 61.72% 62.00% 61.86% Packet delay ms ms ms B. TXOP Continuation performance estimation In order to assess the effect of TXOP continuation, a typical wireless scenario was defined, consisting of 3 QSTAs. The list of TSs hosted within each QSTA is shown in Table IV. The definition of these TSs, as well as the corresponding TSPECs can be found in [6]. Assuming a PHY rate equal to 11Mbps for all TSs, the total aggregated bandwidth value admitted for service (including all transmission overheads) for this scenario was equal to 98%, very near to complete saturation of the wireless medium, while all AC_VI and AC_VO streams were successfully admitted by the QAP. TABLE IV TYPICAL WIRELESS SCENARIO SETUP FOR TXOP CONTINUATION INVESTIGATION QSTA 1 QSTA 2 QSTA 3 Traffic Type Mean data rate (kbps) User Priority AC G.711 Voice AC_VO CD-quality Audio AC_VO G.711 Voice AC_VI CD-quality Audio AC_VI G.711 Voice AC_VO CD-quality Audio AC_VO G.711 Voice AC_VI CD-quality Audio AC_VI SDTV AC_VI Video Conference AC_BK It should be noted that, in practice, the selection of a high PHY rate (11Mbps) is necessary for realizing TXOP continuation, due to the limited allowable TXOP duration (see Table I). More specifically, if PHY is 1Mbps, the transmission of an AC_VO 288-byte packet would require 2.666ms (including all MAC and protocol overheads). Given that the TXOP limit for AC_VO is 3.264ms, no further TXOP continuation is allowed. However, if the PHY rate is set to 11Mbps, only 0.251ms are required to transmit a single packet, hence, provided that the rest 3 rules described in Section III are valid, continuation can be allowed. Table V shows the results obtained when an ideal wireless channel is considered. From this Table, the following results can be obtained: 1. Using TXOP continuation, the measured overall throughput is significantly increased (the mean ThroughputRatio is increased by 26.8%). The additional throughput corresponds to lower priority ACs (AC_VI), as the higher priority AC (AC_VO) originally obtains the

5 resources required to transmit all the packets generated, even when no TXOP continuation is allowed. This can be also verified by the increment of the number of TXOPs granted to AC_VI. 2. An increment of the packet delay is observed when TXOP continuation is applied, due to the extended duration of the TXOPs granted to the winner ACs. This increment is very low for AC_VI and AC_VO, but for AC_BK (which does not employ admission control), it is significantly high. 3. The total number of medium collisions is significantly decreased in the case of TXOP continuation, as the per-packet basis contention and the relative back off operation is decreased. TABLE V MEASURED RESULTS (IDEAL WIRELESS CHANNEL) With TXOP Continuation AC_BK AC_VI AC_VO Mean/Total TXOPs granted Time granted (ms) Throughput (kbps) ThroughputRatio 78.07% 93.81% 99.98% 94.40% Collisions Packet delay (ms) Without TXOP Continuation TXOPs granted Time granted (ms) Throughput (kbps) ThroughputRatio 99.99% 60.86% % 67.60% Collisions Packet delay (ms) The same trends can be observed if the non-ideal channel model with packet loss probability set to 30% is applied (see Table VI). In this case, due to packet retransmissions, the throughput increment with TXOP continuation is lower than in the ideal wireless channel case, but it is still significant. Moreover, the packet delay also slightly increases. TABLE VI MEASURED RESULTS (NON-IDEAL WIRELESS CHANNEL) With TXOP Continuation AC_BK AC_VI AC_VO Mean/Total TXOPs granted Time granted (ms) Throughput (kbps) Convergence 83.62% 76.19% 97.03% 79.49% Collisions Packet delay (ms) Without TXOP Continuation TXOPs granted Time granted (ms) Throughput (kbps) Convergence 99.89% 52.20% 98.06% 60.13% Collisions Packet delay (ms) times and accurate MAC layer modeling, according to the latest e draft. The simulation results are stored in an easy-to-process way and can be exported in many wellestablished file formats. The wireless EDCA simulator accuracy was verified using a number of tests defined by the WFA for certifying interoperable wireless equipment. From these tests, it became obvious that the developed simulation tool efficiently realizes both fundamental properties of EDCA channel access, that is traffic differentiation and AC fairness. Finally, using the wireless EDCA simulator, the TXOP continuation mechanism was analytically investigated. It was found that TXOP continuation significantly improves the overall throughput performance, as the additional transmission time gained due to the reduction of the transmission overheads is granted to lower priority ACs that normally (i.e. without TXOP continuation) demonstrate lower throughput performance. Moreover, TXOP continuation significantly decreases the medium collisions, as the back off-based contention is limited. On the other hand, the main drawback of TXOP continuation is the small increment of the measured packet delay due to the longest TXOPs periods. Especially for high priority ACs, this increment is very low, but for low priority ACs it is very high. Hence, apart from advanced throughput performance, TXOP continuation can be considered as an additional technique for enhancing traffic differentiation. Future plans and extensions of the wireless EDCA simulator include the adaptation to the forthcoming final e specification, as well as the implementation and test of optional e features (such as power save and direct link transmissions). REFERENCES [1] IEEE Std , IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, Nov. 1997, P [2] IEEE WG, IEEE802.11E/D13.0, IEEE Standard for Information Technology - Telecommunications and information exchange between systems Local and metropolitan area networks Specific requirements - Part 11: Wireless Medium Access Control (MAC) and Physical Layer (PHY) specifications: Amendment 7: Medium Access Control (MAC) Quality of Service (QoS) Enhancements, January [3] S. Mangold, S. Choi, G. R. Hiertz, O. Klein, B. Walke, Analysis of IEEE e for QoS Support in Wireless LANs, IEEE Wireless Communications, Vol. 10, pp. 2-12, December [4] A. Floros, T. Karoubalis and S. Koutroubinas, Bringing Quality in the Wireless Arena, Broadband Wireless and WiMax IEC Comprehensive Report, ISBN: x, January [5] A. Floros, T. Karoubalis, CARMEN: Mechanisms for enhancing performance and bandwidth management of Wireless contention based Quality of Service, U.S patent application document, July [6] WiFi Alliance, WFA Interoperability TSPEC List version 0.7, May 24, [7] Wireless Fidelity Alliance, Wireless Multimedia Extension Test Plan draft, Dec VI. CONCLUSIONS In this paper, a novel software tool is presented for simulating the EDCA contention-based access of the e specification. The proposed application realizes a user friendly interface for defining the simulation scenarios, while it provides fast simulation execution