An Enhanced Alternative to the IEEE 802.11e MAC Scheme Aravind Velayutham

J. Morris Chang

Department of Computer Science Iowa State University Ames, IA, 50011 [email protected]

Department of Electrical and Computer Engineering Iowa State University Ames, IA, 50011 [email protected]

Abstract – The original IEEE 802.11 standard has several problems in providing QoS service to stations. The IEEE 802.11e was drafted to overcome these drawbacks. In this paper, we have identified several shortcomings in the components of IEEE 802.11e namely the Enhanced Distributed Coordination Function (EDCF) and the Hybrid Coordination Function (HCF). The EDCF does not guarantee any service to the real-time traffic even though it provides higher priority. Thus we separate the realtime (prioritized) traffic and non-real time (best-effort) traffic and allocate capacity during two different periods namely Contention Free Period and Contention Period. Further we achieve per-flow differentiation by piggybacking future requests in the transmitted packets. This also avoids the overhead of the controlled contention scheme in IEEE 802.11e. The proposed protocol was evaluated by simulations done using ns2 simulator. The performance of the new scheme was compared with that of the 802.11e draft. It was observed that the proposed scheme performs better in giving prioritized allocation as well as increasing the aggregate throughput. Keywords—Wireless LAN, IEEE 802.11e, Quality of Service

I.

INTRODUCTION

Wireless local area networks (WLAN) have come into great use in recent years. There has been tremendous amount of research in this area. The IEEE 802.11[1] was developed as a standard for wireless LANs. There is also an orthogonal increase in the demand for multimedia capabilities of applications. Any network these days should be able to provide Quality of Service (QoS) capabilities to the applications. Thus the idea of providing QoS guarantees over the wireless LANs has been the subject of intense study recently [2]. The IEEE 802.11 standard defined the Point Coordination Function (PCF) to provide reliable service to stations. But several problems with the PCF made the IEEE 802.11 Task group to come with a new version called the IEEE 802.11e [2] which provides QoS to the requesting stations. It includes two new schemes namely the Enhanced Distributed Coordination Function (EDCF) and Hybrid Coordination Function (HCF). In this work, we identify problems with the new scheme and propose enhancements to the 802.11e. We evaluate the performance of the new scheme with the 802.11e scheme and show that the proposed protocol performs better. The rest of this paper is organized as follows. In section 2, we describe the IEEE 802.11 standard. The EDCF and HCF are described in sections 3 and 4 respectively. The problems with the 802.11e scheme are identified in section 5. The

proposed protocol is discussed in section 6. Section 7 describes the simulation results and the paper is summarized in section 8. II.

IEEE 802.11 MEDIUM ACCESS CONTROL

The MAC protocol of 802.11 incorporates two access methods. The basic access method is the DCF which is used to support asynchronous data on a best effort basis. The Distributed Coordination Function (DCF) is based on the Carries Sense Multiple Access scheme with Collision Avoidance (CSMA/CA) protocol. Stations transmit data after detecting that there is no other transmission in progress in the wireless medium. However if two stations detect the channel as free at the same time a collision occurs. The 802.11 defines a collision avoidance procedure to reduce the probability of such collisions. As part of the mechanism, before starting a transmission each station performs a backoff procedure. After detecting the channel as being idle for a minimum duration called DCF Inter Frame Space (DIFS), the station has to keep sensing the channel for an additional random time called the backoff time after which the packet is transmitted. Each station maintains a so-called Contention Window (CW), which is used to determine the number of slots the station has to wait before transmission. The slot interval is defined to be a characteristic property depending on the physical layer. The duration of the backoff time is determined as a random interval from 0 to CW. All stations have equal probability to access the channel and thus share it equally. But this method has no guarantees for queuing delays so it is not optimal for time-bounded applications. The contention free service for time-bounded traffic is provided by the Point Coordination Function (PCF) which basically implements a “polling” access method. This scheme lets stations have priority access to the wireless medium. The PCF uses a point coordinator (PC) usually at the access point which periodically polls stations giving them the opportunity to transmit frames and thus avoiding any contention for the channel. The PCF has higher priority than the DCF because it may start transmissions after a shorter duration (PCF Inter Frame Space) PIFS than the DIFS. With PCF, a Contention Free Period (CFP) and a Contention Period (CP) alternate over time in which a CFP and the following CP form a superframe. During the CFP, the point coordinator polls the station for transmitting their frames while during the CP the stations contend among themselves using the DCF mechanism. It is mandatory that a superframe includes a CP of a minimum length that allows at least one data packet delivery under DCF. The point coordinator generates a beacon frame at

regular beacon frame intervals called target beacon transmission time (TBTT) and the value of the TBTT is announced in the beacon frame. After sending the beacon frame, the point coordinator periodically polls the stations to transmit frames and ends this CFP with a CF-End control frame. There are several problems [3] with the PCF that led to the development of enhancements to the protocol. The problems of the PCF scheme are as follows : !

The transmission of the beacon by the point coordinator depends on whether the medium is idle at the time of TBTT. But after the medium is idle the PC will get priority due to shorter PIFS. But the time at which the medium becomes idle is unpredictable. Thus the beacon frame can get delayed affecting the time allocated to time-bounded traffic. This is referred to as the deferred beacon problem of PCF. ! Further the duration of the transmission that happens after the polling is not under the control of the point coordinator. IEEE 802.11 task group has defined a few enhancements to the PCF and the new MAC scheme is called 802.11e which introduces the Enhanced DCF (EDCF) and Hybrid Coordination Function (HCF). III. ENHANCED DISTRIBUTED COORDINATION FUNCTION (EDCF) The various streams are classified into Traffic Categories (TCs). EDCF uses different mechanisms to provide service differentiation. The minimum contention window for the backoff is different for different priority classes. This will in turn reflect on the higher priority classes getting more transmission time than lower priority classes. Further different inter frame spaces can be used for different priority classes. The EDCF is operative only during the Contention Period (CP). During the CP, each TC within the stations contends for a Transmission Opportunity (TXOP) independently. Each TC starts a backoff after detecting channel to be idle for a time interval equal to Arbitration Inter Frame Space (AIFS). The value for AIFS is dependent on the traffic category the traffic belongs to. The backoff is set to a counter which is a random number from the interval [1, CW+1]. As in DIFS contention for each collision of the frames the CW value is increased. The initial value of CW is set to CWmin. The value of CWmin is also dependent on the traffic category of the stream. After a collision is detected the CW is increased as follows newCW [TC] = (oldCW [TC] +1) * PF [TC] – 1 Here PF is the Persistence Factor which is also a traffic category dependent parameter. PF determines the degree of increase of the Contention Window when collisions occur. Higher priority traffic will have lesser PF value than lower priority traffic PF value. Thus when collisions occur the higher priority traffic flows’ CW value will increase by a lesser value than a lower priority traffic’s CW value.

IV. HYBRID COORDINATION FUNCTION (HCF) The polling scheme of PCF is extended in 802.11e by using the Hybrid Coordination Function (HCF). In this scheme, there is a hybrid coordinator (HC) usually co-located with the access point1. The HC may allocate TXOPs to itself to initiate frame transmission after waiting for a time equal to PIFS which is shorter than DIFS and any AIFS. Thus the HC gets priority over other nodes to transmit frames. The HCF is operative during both the CP and CFP durations. During the CP each station gets its TXOP either when the medium is determined to be available under the EDCF rules or when the station receives a QoS CF-Poll frame from the HC. During the CFP, the starting time and maximum duration of each TXOP is specified by the HC using CF-Poll frames. As the name (contention free period) denotes, stations cannot contend among themselves for TXOP during the CFP. The CFP ends either at the time specified in the beacon frame or by a CFEnd frame sent by the HC. The 802.11e also uses another mechanism by which the stations send update information to the HC. This includes which stations need to be polled, polling time and duration of transmissions. The mechanism used is called controlled contention in which the HC allocates a number of controlled contention opportunities separated by SIFS. This is done so that stations with high priority traffic need not contend with other EDCF traffic for transmitting the request information. The HC also sends out a filtering mask containing the TCs in which resource requests may be placed. Each station chooses one opportunity interval and transmits a resource request frame containing the requested TC and TXOP duration. The HC also sends out an acknowledgment control frame so that requesting stations can detect collisions during controlled contention. V. PROBLEMS WITH THE 802.11E We have identified several drawbacks in the proposed schemes and have developed enhancements to the problems. First we describe the problems that are present in the 802.11e scheme. A. Controlled contention scheme The HCF uses the mechanism of controlled contention for updating service information for each station at the HC. This mechanism helps the stations with priority traffic to send their requests for channel allocation (request for polled TXOPs) to the HC. Thus HCF allocates a separate time interval for the transmission of traffic information from the stations with priority traffic to the HC. Further the stations can update the channel requirement information only when they get the opportunity during the contention phase. Thus this is a passive process where a change in allocation requirement cannot be transmitted immediately. B. Per-station priority During the controlled contention each station gets an opportunity to send out a resource request frame. But this 1

In this paper we have used the terms Access Point, Point Coordinator and Hybrid Coordinator interchangeably

service information received by the HC is per-station. Thus all flows from a station must have the same priority level. This is not necessarily true in real life applications. There might be applications with different requirements running in the same station. Hence if we chose a particular priority level for a station then lower priority flows will get more allocation than is required and requirements of higher priority traffic will not be satisfied. The per-station priority model also leads to another problem. Consider TCP flows from the stations to the HC. The HC which has a single priority can transmit the TCP_ACKs back to the stations only at a fixed rate which is independent of the priority of the TCP flows. This decreases the throughput of the TCP flows at the stations. This is because in TCP during congestion avoidance phase, a source waits for a new ACK before generating a new packet. Thus if the HC is not sending the ACKs back at the specific priority rate, the throughput of TCP flows suffer at the stations. This phenomenon was also reported by the authors in [4]. Thus the downlink flows form the HC to the stations should have different priorities based on their requirements. C. Problems with EDCF The differentiation mechanisms used during EDCF phase in 802.11e gives priority based on traffic categories. EDCF does not give any guarantees for priority traffic. At high loads, there are a high number of collisions even for flows with high priority. Even though the average delay is small for high priority traffic, at high loads the packets have very long delays. As shown later, the simulation results prove this fact. VI. PROPOSED SCHEME We describe the details of the proposed model in this section. In the new protocol, each station maintains separate queues for different traffic categories as in the EDCF mechanism of 802.11e. But these queues are also used to notify the HC about the required service for the flows from the station. We use an alternative approach to transmit channel requirement information from the stations to the HC. In this approach each packet is attached with the request for future allocation. Thus while transmitting a packet each flow has a chance to request for future channel allocation. The HC uses this information to identify and maintain details about flows that need priority. Thus we are able to achieve per-flow service differentiation. Further by piggybacking the request for future allocation in the packets, this model avoids a separate controlled contention phase to transmit service information to the HC. This model works even in the case of direct link access between two stations. A protocol namely the Direct Link protocol has been proposed as an addition to the IEEE 802.11e, as illustrated in Fig.1. This protocol allows the transmission of frames directly from one station to the other without going through the access point. This protocol is useful in cases where a station in power-save mode can be woken up by the access point and then there can be direct transmission to the station from another station. In this case, the sending

Figure 1. Direct Link Protocol of IEEE 802.11e draft

station will send the request for a slot in the next schedule in the transmitted packet itself. The HC also has information about the service requirements of the downlink traffic. Thus downlink traffic is also treated in the same way as the upstream traffic. So perflow differentiation is also applied for packets from the HC to the stations. This will be particularly useful in cases where the TCP data flows from the stations to the HC and the ACKs flow from the HC to the stations. Thus both the upstream TCP data and downstream TCP-ACK flows are having the specified priority. To piggyback the request in the frame transmitted from the station, we use the More Data field of the frame control field, as shown in Fig.2.. This is usually used by the AP to indicate to a station in power-save mode that more frames are buffered for the station at the HC. In the original 802.11 frame format, the More Data field is valid in frames transmitted by HC to stations. A value of 1 indicates that at least one additional buffered frame is present for the same station. The More Data field may be set to 1 in frames transmitted by stations during the CFP to indicate to the HC of buffered frames. In our scheme we use the More Data field to indicate that the specific flow needs to be polled during the next schedule. The frames transmitted during the CP also can have the More Data bit set to 1, indicating to the HC to include that flow also in the poling list from now. From the information it has collected about the channel allocation requirements of various flows in the network, the HC uses a scheduling mechanism to come up with a polling schedule for each Contention Free Period. The transmission schedule is illustrated in Fig.3.

Figure 2. Frame Control Field of IEEE 802.11 standard

intervals. This is better because the performance of that scheme depends on the relative values of DIFS intervals as well as persistence factor and CWmin. Any new stations that join the Basic Service Segment (BSS) of the HC will issue, along with the association request, any requirements for priority traffic. This is used by the HC in future scheduling. Even stations that want to change their status from non-real time traffic to real time traffic can attach a request in the packet sent during the contention period. This will notify the HC to include the station in the polling list. VII. SIMULATION RESULTS Figure 3. Transmission Schedule of the proposed scheme

A control frame called the CFP-Schedule is transmitted at the start of each CFP after the transmission of the beacon frame. The CFP-Schedule frame is a management frame similar to the beacon frame and contains the stations and their TXOPs. This is similar to the TXOPs in HCF. The CFPSchedule also indirectly specifies the end of the CF period. It is also to be noted that the downstream priority traffic also gets a transmission slot in the CFP-Schedule. Further in the proposed model there is no notion of EDCF. The EDCF provides no guarantees to the real-time traffic. It just gives them more priority. Since in our model, the stations with real-time traffic receive their allocation during the CFP, only the stations with best-effort traffic get transmission duration during the Contention period. This period is exclusively used for the non-real time traffic which needs best-effort service. Thus at the end of the CFP the HC sends out another control frame called the CP-Schedule which just specifies the stations that can contend for transmission during the Contention Period. The scheduling of the traffic is done by the HC using various possible techniques. One of the possible scheduling schemes is to give maximum allocation to real time traffic and give a specified minimum allocation to non-real time traffic. The delay perceived by the real-traffic is in accordance with their Traffic Category and non-real-time traffic is given besteffort service.

We analyzed the performance of our protocol by simulating experiments using ns-2 [6]. A simple simulation set up with three stations and an access point is used, as shown in Fig.4. We have compared the performance of our scheme with that of the 802.11e. We call our proposed scheme E-802.11 for Enhanced 802.11. In our simulations flows, a flow with priority level 1 has greater priority than flows with priority 2. Thus priority is inversely proportional to the priority level. A. Scenario I In this case we simulate three different priority UDP flows from the stations to the access point. We have analyzed the average throughput obtained by the three flows using both the 802.11e scheme and E-802.11. As seen in Fig.5, the different flows get throughputs according to their priorities. But the channel utilization is better in E-802.11. This is seen from the increased aggregate throughput in both the schemes. This is attributed to the fact that there is lesser overhead in E-802.11 due to piggybacked allocation requests. We can see in Fig.6 that the drop rates for the flows are also better in E-802.11 than in 802.11e. B. Scenario II In this case we simulate three TCP flows with different priorities. As seen from Fig.7, using 802.11e there is no considerable prioritization effect among the three flows. But we see that E-802.11 achieves significant prioritization. The three TCP flows get throughput in accordance with their

The CFP-Schedule that the HC sends just at the start of the Contention Free period contains the Transmission Opportunity (TXOP) for each station that has attached real-time traffic service requirements in the previous packets. By specifying the exact transmission duration for each station, this scheme solves the problem of uncertain transmission time taken by stations using the HCF scheme. Further the HC can ensure the accurate allocation of bandwidth to various real-time traffic categories. The HC also starts the contention period explicitly by sending the CP-Schedule. During this contention period the stations, more specifically flows that are specified in the CPSchedule frame can transmit frames after normal DCF contention mechanisms. Thus we avoid the differentiation mechanism using varying DIFS (AIFS in EDCF) and backoff

Figure 4. Simulation setup

C. Scenario III Here we try to analyze the case where two stations have the same priority. So we simulate a scenario where two UDP flows have the same priority and the third has a different priority. In Fig.7, we see that E-802.11 achieves the same throughput for the UDP flows with the same priority. This is due to the exclusive CFP free period access for priority traffic. Thus the priority traffic has guarantees with respect to the allocation of channel.

0.79 Average Througput (Mbps)

priority using E-802.11. This is due to the per-flow differentiation achieved by E-802.11 instead per-station fairness as in 802.11e.

0.785 0.78 0.775

UDP priority 2

0.77

UDP priority 1

0.765

UDP priority 1

0.76 0.755 0.75 0.745 2

3

4

5

6

7

8

9

10

Offered Load (Mbps)

Figure 7. Throughput of UDP flows with varied offered loads

UDP Priority 1 Throughput (Mb)

3.5 3

UDP priority 2

2.5 2 1.5

UDP priority 3

1 0.5 0 802.11 e

E-802.11

Aggregate Throughput

Figure 4. Average Throughput using UDP with priorities

Throughput (Mb)

1.2 1 0.8

SUMMARY

We have identified several drawbacks of the IEEE 802.11e scheme that has been drafted for providing QoS in wireless LANs. We have designed a new protocol that overcomes the problems of 802.11e. The new scheme removes the EDCF used in IEEE 802.11e. This is because it was found that the real-time traffic is not guaranteed any allocation during EDCF. It is just that they receive priority. So instead of using the EDCF, the proposed protocol uses a complete polling scheme for real-time traffic and contention access for besteffort (non-real time) traffic. The HC uses a scheduling mechanism to determine the length of the contention free and contention period. The new protocol uses two management frames namely CFP-Schedule and CP-Schedule to transmit the TXOPs from the HC to the stations. The performance of the proposed scheme was found to be much better than the IEEE 802.11e scheme.

TCP priority 1

0.6

TCP priority 2

0.4

TCP priority 3

0.2 0 802.11e

E-802.11

Figure 5. Average Drop Rate using UDP with priorities

80

Drop Rate Percentage

VIII

70 60 50

UDP Priority 1

40

UDP Priority 2

30

UDP Priority 3

20 10 0 802.11e

E-802.11

Figure 6. Average Throughput using TCP with priorities

REFERENCES [1] "Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications," IEEE 802.11 standard, 1997. [2] Heegard, C. and Coffey, J. and Gummadi, S. and Murphy, P. A. and Provencio, R. and Rossin, E. J. and Schrum, S. and Shoemake, M. B., “HighPerformance Wireless Ethernet”, IEEE Comm. Magazine, vol. 39, no. 11, Nov. 2001. [3] IEEE 802.11 WG, Draft Supplement to STANDARD FOR Telecommunications and Information Exchange Between Systems - LAN/MAN Specific Requirements - Part 11: Wireless Medium Access Control (MAC) and physical layer (PHY) specifications: Medium Access Control (MAC) Enhancements for Quality of Service (QoS), IEEE Std 802.11e/D4.0, Nov. 2002. [4] S. Mangold, S. Choi, P. May, O. Klein, G. Hiertz, L. Stibor. “IEEE 802.11e Wireless LAN for Quality of Service”, Proceedings of the European Wireless Conference, Florence, Italy, February 2002. [5] I. Aad and C. Castelluccia. Differentiation mechanisms for IEEE 802.11. In Proceedings of INFOCOM, 2001. [6] NS-2 simulation tool home page. http://www.isi.edu/nsnam/ns/, 2000