Admission Control Scheme for QoS Support in IEEE 802.11e Wireless LAN

INTRODUCTION

With the increased requirements of streaming applications, the facilities to provide QoS become vital in WLAN. The conventional Distributed Coordination Functions (DCF) of the wireless standard IEEE 802.11 provide only the best-effort services (Bianchi, 2000), hence QoS is not guaranteed (Ni et al., 2004). The IEEE 802.11 task group E thus initiated the new scheme of channel access known as EDCA in the IEEE 802.11e standard to provide better channel performance than the best_effort services by guaranteeing the QoS parameters like throughput (Zhu and Fapojuwo, 2007), delay and jitter (Banchs et al., 2002; Huang and Liao, 2007). Comparing to DCF, the EDCA standard provide differentiated service by prioritization. EDCA uses four access categories (ACs) for providing prioritized QoS. These ACs are defined as AC_BK (BackGround), AC_BE (Best Effort), AC_VI (Video) and AC_VO (Voice), respectively as four mechanisms for channel accesses. Each access categories maintains its queue and a set of access categories factors.

AIFSN: Arbitration Inter Frame Space Number is the channel access parameter for the transmission of packets. It is the minimum time interval between the wireless channels becoming idle to begin the transmission of packets.

CW: Contention Window-a back_off mechanism, when more than one station contends to access the channel. It is a random number.

TXOPLimit: It is the maximum time interval for which a QoS station (QSTA) can transmit the packet after obtaining the transmission opportunity.

When each access category contends to access the channel they must defer for an waiting period AIFS[AC]:

AIFS[AC] = AIFSN[AC].aSlotTime+aSIFSTime

where, the parameter aSlotTime and aSIFSTime is the short interframe space time period. AIFSN is used to determine the length of AIFS. This begins the back_off counter to delay the channel access a CW for a random period. In addition to the Inter Frame space differentiation, certain CW lengths are specified for different ACs. Access Categories are differentiated based on their priorities by using set parameter like AIFS[AC], CWmin[AC] and CWmax[AC]. An AC with shorter CW is able to transmit the frame ahead of low priority AC. The higher priorities ACs have smaller CW, CW_min and low priority ACs with larger CW, CW_max size.

For ACi (i = 0, 1,.. 3) the initial CW size is CW_min[i] and maximum CW size CW_max[i].

In this study, we analyzed that, by optimizing contention window value results to improve the WLAN throughput. Considering this characteristics of a wireless channel we introduce an admission control mechanisms (Jamin et al., 1996) that adapts CW based on the channel condition and adjust the CW value to maximize the throughput for the QoS station (QSTA). The admission control algorithm (Jamin et al., 1996; Achary et al., 2012a) will accept the station (STA) to access the channel.

The rest of the study is organized as follows. In the next section we provided the details about EDCA and CW which is followed with the throughput analysis based on the CW parameters and the proposed admission control algorithm. The performance evaluation and the conclusions are at the end of the study.

ENHANCED DISTRIBUTED CHANNEL ACCESS (EDCA) AND CONTENTION WINDOW (CW)

IEEE 802.11e uses EDCA as one of the preferred channel access method. It uses service differentiation (Achary et al., 2012b) The enhanced distributed channel access defines four Access Categories (ACs) to provide priority based services with CW size, AIFS value and TXOP for a specific AC to provide the required QoS. Each access category ACi, has its own priority value (i = 0, 1, 2, 3), where 0-BE(AC_0), 1-BG(AC_1), 2-VI(AC_2) and 3-VO(AC_3).

The priority levels of an AC are determined based on AIFS and CW values. The smaller AIFS and CWmin values or larger TXOP for an AC represents that, it has a higher priority value than the other ACs. i.e., an AC with higher priority value has smaller CW (CW_min). The CW value ranges from CW_min to CW_max. The transmission of a frame is based on the EDCA back_off mechanism. For transmission of a frame the STAs wait for a period of AIFS[AC] and then execute a back_off procedure by setting the back_off counter to a random number determined from CW value range (1, CW+1). The CW is initially set to CW_min. The back_off counter value is reduced in every slotTime When the counter reaches to minimum value the back_off procedure transmits the packet. If two or more STAs attain the CW_min value at the same time slot a collision occurs. Once the packets are received successfully and ACK (acknowledgment) is sent by the receiving STA and waits for a small time interval say SIFS (short inter frame space). This acknowledgment indicates that the receiving STA has successfully received the packets. If there is a collision a new CW value CW_new is calculated according to CW_new [AC] = [CW_current (AC)+1xPF(AC)] the Persistence Factor (PF) value decreases as the priority increases. EDCA traffic parameters are given in Table 1 and the values of CW(min), CW(max), AIFS and TXOP for different Ac’s are mentioned in Table 2.

Figure 1 shows the service differentiation (Achary et al., 2012a) mechanism introduced by EDCA by assigning appropriate priority values (CW_min, CW_max and AIFS) to the ACs. The AC with lower values (CW_min, CW_max and AIFS) has the greater probability of accessing the wireless channel. Individual access category acts as a virtual STA contends to access the wireless channel and independently starts its back_off procedure after finding the channel is idle for a period of AIFS. Once the AC gain access to the channel, the time period for which a particular STA can begin the transmission called TXOP. It is a time period in which the STAs are allowed to transmit multiple packets from the same AC. In IEEE 802.11e EDCA, ACs with higher priority values will have longer TXOP time interval due to smaller (CW_min, CW_max and AIFS).

Table 1:	EDCA traffic priorities mapped to AC

Table 2:	EDCA parameters for different AC’s

Fig. 1:	4 ACs in EDCA

Back_off contention algorithm: Carrier sense multiple access collision avoidance (CSMA/CA) is the channel access mechanism used in IEEE 802.11 to access the medium through a random back_off value. The medium can be accessed if it is free for a period of DIFS or less than that, else it will enter into the contention phase. Then the STA choose a random back_off time within the CW and delays the channel access for this random time. The finite state machine for binary exponential back_off is as shown in the Fig. 1 above. The transmission process and its interframe timing in the medium access mechanism of IEEE 802.11 protocol is given in Fig. 2. The arrow in the Fig. 3, represents the switching of the back_off algorithm for one state to another. A cause for this switching is indicated above the line and the results out of this are shown below the line. In Fig. 3, the state transition occurs based on the event if no action is take on transition then the symbol Λ is used below the horizontal line. Whenever the packet is ready to send, the STA performs carrier sensing and goes to the next state in which it waits for the amount of duration given by Waiting Time (WT). Once the waiting time = Rand (0, CW)xslotTime it goes to the transmission state. There are three possible events at this state; as If there are no more packs to send then the STA go to idle state. When a collision occurs during transmission CW is adjusted to CW = min (2CW, CW_max ) based on the back_off algorithm used. Finally the STA goes to waiting STA to wait for the newly calculated waiting time.

Fig. 2:	Medium Access mechanism of 802.11 (Pereira et al., 2007)

Fig. 3:	CW State mechanism

THROUGH ANALYSIS AND CW

To improve the performance of the WLAN, IEEE 802.11e working group has proposed admission control (Achary et al., 2012b). The two entities play a major role in this admission control are; the QSTA that implement the QoS and QAP (Karanam et.al., 2006) that support the QoS.

The proposed algorithm provides performance guarantee (Huang and Liao, 2007) based on the centralized access point as in Fig. 4, the quality of service access point (QAP). A simple admission control procedure (Achary et al., 2012a) is as follows; when a QSTA requests the desired QoS to the QAP which process to decide, whether the request can be granted. As a response to the request form a QSTA, the QAP sends a response message to QSTA including information of acceptance or rejection. In case of acceptance the QAP computes the new configuration for the accepted QSTA, with a guaranteed throughput and delay (Banchs et al., 2002) and then distribute the processed required parameters to the STAs, otherwise it is rejected. The accepted STAs are allowed to send the packets.

Admission control is one of the efficient mechanisms to provide QoS using resource management technique. The WC_min is one of the significant QoS factors that greatly affect throughput and transmission delay (Banchs et al., 2002; Huang and Liao, 2007), for the packets of smaller size. In this segment we analyze the QoS of IEEE 802.11e WLAN, with respect to its transmission delay and performance (Ho et al., 2007) by changing the CW_min value.

Throughput analysis: Consider a WLAN with ‘n’ number of QSTA contending to access the wireless channel. The QAP calculate the CW_min parameter for each accepted QSTA and also for the new requesting QSTA; which satisfy the transmission delay and throughput requirements.

Assume that if QSTA i with a CW, CW_i to access the channel. The probability (Ho et al., 2007) that the QSTA i transmits the frame is:

(1)

where, i = 1, 2,……n.

Fig. 4:	QoS Architecture

And the probability of the frames which are transmitted successfully from a QSTA i is:

(2)

If:

(3)

Then the probability of successful frame transmission Ps from Eq. 2 and 3 is given by:

(4)

The availability of the empty time slot in a given channel is given by:

(5)

The collective performance is computed by Eq. 2:

(6)

Where:

Using Eq. 4-5 the probability for packet collision is obtained as:

(7)

Substituting Eq. 7 in 6 we can write Eq. 6 as:

(8)

To find the number of accepted QSTAs, consider the relation between the throughput of QSTAs, QSTA i and QSTA j as:

(9)

(10)

where, τ_i is very small.

This ratio in Eq. 10 is more accurate as the number of accepted QSTAs are very large. When the number of QSTAs almost saturates the channel, the QAP starts rejecting the STAs.

From Eq. 10, it follows:

(11)

The throughput of QSTAi, r_i is:

(12)

Throughput analysis based on CW: Let the throughput used by the QSTAi is r_i as in Eq. 12 and the throughput requirements of the station is, Ri.

The throughput requirement of the entire CW set {CW₁, WC₂,……………, CW_n} for ‘n’ QSTAs is ∀iε{1, 2, 3……………, n}r_i≥R_i.

The QAP compute the CWmin values ranging from {CW₁, WC₂,……………, CW_n} for the accepted QSTA i and also for the contending QSTAs, it satisfy the performance requirement set {R₁, R₂,……, R_n} and delay bound. This range is considered as critical range. To improve the QoS (Ho et al., 2007) of the IEEE 802.11e WLAN, admission control algorithm in the QAP should accept as many as possible to optimize the critical range with r_j/R_j ratio minimum:

(13)

Using Eq. 10 and 13:

(14)

In Eq. 12:

(15)

If L, Ts and Te are constants, the resulting equation will be:

(16)

Maximizing Eq. 16 will maximizes all:

(17)

When ‘n’ is very large τ₁<<1:

(18)

Where:

(19)

Optimization of the τ₁, τ₁* values, for maximizing is by:

(20)

The optimal CW set that maximizes all r_i’ is ∀i:

(21)

From the above expression we can infer the relationship between CW_i and the throughout. i.e., the contention window CW_i is inversely proportional to the throughout.

ADMISSION CONTROL ALGORITHM (ACA) USING CW

In this algorithm for the proposed admission control mechanism (Jamin et al., 1996; Achary et al., 2012b) the WLAN is working with a set of CW, {CW₁, CW₂,…………, CW_n} that fulfills the required performance of the number of (n) QSTAs:

Here, the throughput experienced by a QSTAi Ri. If an additional station, STA(n+1) is added to the set with a throughput requirement (R_n+1) the QAP first compute a new CW set. For an IEEE 802.11 wireless system in (Tay and Chuo, 2001) proposed the following equation which dynamically adjusts the value of CW_min considering the number of (n) existing QSTAs.

(22)

Where:

(23)

TAIFS[i] is the time period of arbitration interframe space (AIFS) for AC[i], 0≤i≤3.

In IEEE 802.11e the four ACs consider this adjustment factor (K) to prioritize the ACs.

The new CW set is {CW₁', CW₂',……………, CW_n', CW_n+1'} using Eq. 21.

Then the throughout that would be received by the n+1, QSTA with this new CW set will be calculated using expression 12.

If the resulting throughput meet the requirements ∀iε{1, 2, 3,……………, n, n+1} r_i’≥R_i.

The STA (n+1) is considered to access the wireless channel and the new CW set is circulated to all the STA’s. Else it will not be considered to access the channel. In the proposed admission control algorithm the QSTA sends ADD Traffic Stream (ADDTS) request to the QAP before starting the real time data flow. This ADDTS includes the traffic priority and traffic specification (TSPEC).TSPEC specifies QoS requirements in terms of throughput. The QAP after receiving the ADDTS information checks the QoS requirements and decides the existing resources of WLAN could house the data frames. If sufficient resources are available the frame is accepted, else the QAP evaluate the effect of these newly admitted data frames and then reject the new request, otherwise these data frames may down grade the performance of WLAN.

PERFORMANCE ANALYSIS

The simulation model for the admission control scheme (Achary et al., 2012a) in this paper is developed using NS-2. Simulation performed using eight, IEEE 802.11 QSTA with EDCA (Kong et al., 2004; Xiang et al., 2007) channel access mechanisms were configured into infrastructure mode are used. The highest data rate of the physical layer used here to simulate the wireless medium is up to 11 Mbps. The MAC layer is modified to support IEEE 802.11e, EDCA. Data frames of all the four traffic classes are fed in to the corresponding ACs; AC[0], AC[1], AC[2] and AC[3], respectively. All the sources in this are the Constant Bit Rate (CBR) sources over UDP. The size of the packets to be transmitted is 1500 bytes and the data rate of the node is 600 Kbps. Simulation parameters used here are summarized in the Table 3. In conventional IEEE standards line IEEE 802.11b, as the number of STAs increases, sharing of network resources also increases, the jitter and reduces the throughput by readjusting the CW value. This results no guaranteed QoS for streaming traffic.

Table 3:	IEEE 802.11 WLAN Simulation parameters value

The performance (Ho et al., 2007) of IEEE 802.11e WLAN using EDCA (Kong et al., 2004; Xiang et al., 2007) technique in MAC layer shows a more stable performance with guaranteed bit rate based on the prioritized packet delivery process (Bianchi et al., 1996). Each station is configured to provide a minimum value of performance; thus higher priority data frames are protected against those of the data frames with lower priorities. Figure 5 shows the decrease in packet rejection rate which indicates that in IEEE 802.11e EDCA with Admission Control Algorithm (ACA) experiences lower packet drop rate. The Fig. 6 represents end-to-end delay.

Fig. 5:	Number of dropped packets w.r.t number of connections

Fig. 6:	Normal end-to-end packet transmission delay w.r.t number of connections

Fig. 7:	Average normalized throughput w.r.t number of connections

In which packets with higher priorities are less degraded by end-to-end transmission delay. In general the simulation result in Fig. 7-9 shows the proposed admission control algorithm exhibit an increased bit rate, when the wireless channel traffic is increased.

Fig. 8:	Percentage channel utilization w.r.t number of connections

Fig. 9:	Normalized overall throughput w.r.t number of connections