This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE CCNC 2010 proceedings ㆍSense channel during DIFS Backoff Backoff Backoff counter resume counter resume counterp DIFS Contention Window paused paused aused DIFS PIFS Backoff- Data Busy Medium SIFS Next Frame Window transmission Node A Slot time Defer Access ㆍBackoff slot reduced when channel is idle Node BFigure 1. Illustration of the IEEE 802.11 DCF mechanism Node C The BEB degrades the performance of the network when Node Dthe network is heavily loaded because each new packet starts timewith the minimum CW. This resetting behavior becomes veryunstable when numerous nodes are contending within the Figure 2. Estimating the number of active nodes with the PCBsame wireless channel. This can cause more collisions and itdecreases the whole system’s utilization. Fig. 1 shows how the that is using the wireless channel, and so the traffic load of theDCF works. network is determined by the number of pauses. The PCB counts the pauses during the countdown procedure and it setsB. The EIED and the EILD an appropriate CW size for the current traffic load of the In the EIED , the CW exponentially increases by a network. Fig. 2 describes the PCB.backoff factor of rI whenever a collision occurs, and itexponentially decreases by a backoff factor of rD if a node D. The HBABsuccessfully transmits a packet. The EIED can be given as. The HBAB algorithm checks the last N states of the medium (N=2 in this implementation), and it determines ⎛Transmission success : CW = CWold / rD ⎞ ⎟ whether to increment or decrement the CW value based on the ⎜ ⎜ ⎟ ⎜Transmission fail : CW = CWold × rI ⎟ ⎟ (2) channels tendency to being free or busy . The HBAB ⎝ ⎠ algorithm fixes two parameters, α and β, which are used to , ( rI > 1 and rD > 1). increase or decrease the new CW based on the old CW value. The EILD linearly decreases by a backoff factor of rD. The TABLE 1 shows the suggested CW values per state check (0EILD can be expressed as follows: indicates both a busy channel and 1 indicates a free channel. ⎛Transmission success : CW = CWold − rD ⎞ ⎟. III. THE PROPOSED BACKOFF ALGORITHM ⎜ ⎜ ⎟ (3) ⎜Transmission fail : CW = CWold × rI ⎟ ⎟ ⎝ ⎠ The proposed algorithm has two main functions: The estimation scheme for the number of active nodes and the The EIED and the EILD methods are based on partial optimal CW allocation scheme are shown in TABLE 2. Theobservations, such as that each node uses its own results of estimation scheme exploits the number of idle slots in thetransmissions to represent the whole system. The results of backoff period in order to derive the exact number of activeboth the transmissions and the system load may have a positive nodes. The optimal CW allocation scheme uses the estimatedcorrelation, but they are not sufficient to precisely set the CW number of active users in order to enhance the systemvalue. performance. The detailed description is as follows.C. The PCB A. Estimating the number of active nodes The PCB monitors the traffic load of the network, and the In step 1 in Table 2, each node obtains the average numberPCB sets an appropriate CW to match the traffic load of the of both the idle slots and the busy slots during the backoffnetwork . The countdown procedure in the backoff period period. Given N slots in the total backoff period and n nodes,pauses when other nodes simultaneously use the wireless the probability that r out of n nodes transmit their data during achannel. Therefore, each pause represents more than one node slot is given by ⎛ n⎞⎛ 1 ⎞ ⎛ r n-r TABLE I. THE CW ESTIMATION ALGORITHM IN THE HBAB ⎞ 1 P( X = r ) = ⎜ ⎟ ⎜ ⎟ ⎜1 − ⎟ . (4) Ex: CW value ⎝ r ⎠⎝ N ⎠ ⎝ N⎠ State CW value (with α=1 β=2) The number r in a particular slot is called the occupancy number of the slot . The expected number of slots, with the 00 CW=CWold × (α β) 2 CWold occupancy number r, is given by 01 CW=CWold × (α / β) 1/2 CWold ⎛ n⎞⎛ 1 ⎞ ⎛ r n−r 1⎞ 10 CW=CWold × (β / α) 2 CWold E[ X = r ] = N ⎜ ⎟ ⎜ ⎟ ⎜1 − ⎟ . (5) ⎝ r ⎠⎝ N ⎠ ⎝ N⎠ 11 CW=CWold × (1/ α β) 1/2 CWold To estimate the number of nodes (nest), this paper defines the average number of idle slots a0(N, n), which means the ratio
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE CCNC 2010 proceedingsof the number of the idle slots to the number of slots in the 1 ⎛ 1⎞ n −1backoff period is given by Psucc , N = ×⎜1 − ⎟ ⎜ ⎟ ×N N ⎜ ⎝ N⎟ ⎠ (10) n n −1 ⎛ 1⎞ ⎛ 1⎞ = ⎜1 − ⎟ a0 ( N , n) = N × E[ X = 0] = N × ⎜ 1 − ⎟ . (6) ⎜ ⎟ . ⎝ N⎠ ⎜ ⎝ N⎟ ⎠ By using (6), the number of users can be derived as Let Psucc(k) be the probability that a node successfully transmits a frame in the kth retransmission. Then Psucc(k) is log(a0 ( N , n)) − log( N ) nest = . (7) Psucc (k ) = Psucc , N (1− Psucc , N )k −1 . (11) log( N −1) − log( N ) Thus, the average number of retransmissions is After the end of the backoff period, a node can calculate thetotal backoff period N and the estimated number of active users, ∞ 1as shown in TABLE 2. E ( X = k ) = ∑ kPsucc ( k ) = n −1 . ⎛ 1⎟⎞ (12) k =1 ⎜1 − ⎟ ⎜ ⎜ ⎝ N⎟⎠B. Deciding the optional CW This paper derives the optimal CW based on the average Therefore, D(N, n) can be obtained from (8) and (12) asaccess delay D(N, n) which refers to the time that is needed to Ntransmit a packet from one node to the other. D(N, n) can be D ( N , n) = n −1 . ⎛ ⎞ (13)obtained as follow . ⎜1 − 1 ⎟ ⎜ ⎟ ⎜ ⎝ N⎟⎠ D( N , n) = number of retransmission × total backoff size. (8) In (13), D(N, n) depends on N and n. Since N is the The probability that a node successfully transmits its data system’s parameter, this paper drives the optimal N toduring a slot is given by minimize D(N, n). Since D(N, n) is a concave function with n−1 respect to N, the optimal N can be obtained by differentiating 1 ⎛ 1⎞ D(N, n) with respect to N as Psucc = ×⎜1− ⎟ ⎜ ⎟ , (9) N ⎜ N⎟ ⎝ ⎠ ∂ ∂ N D( N , n) = n −1 = 0.where 1/N is the probability that a node transmits its data at the ∂N ∂N ⎛ ⎞ ⎜1 − 1 ⎟ ⎜ ⎟ (14)particular slot in a backoff slot. Based on (9), the probability ⎜ ⎝ N⎟⎠that a node successfully transmits a frame during the totalbackoff period is given by From (14), the optimal CW can be obtained as CWoptimal = n . (15) TABLE II. THE EBA ALGORITHM IV. THE SIMULATION RESULTSStep1: Estimating the number of active nodes This section evaluates the system performance in terms of the throughput and the average access delay. This paper When a channel is busy during the backoff period simulates the IEEE 802.11b based WLAN setup module as -. busy_count = busy_count +1 defined in the OPNET. The range of the number of nodes is Backoff period end within 30 ~ 70 and the simulation time is 300 seconds. All nodes are within one hop distance and they select a random Calculate the parameters -. busy_slot_count=busy_count * α, destination. The parameters that were used in the simulation are listed in Table 3. The parameters rI and rD in the EIED are set ⎛ data _ packet _ size 1 ⎞ ⎟ to 2, as suggested in . ⎜ ⎜α = × ⎟ ⎜ ⎟ ⎟ ⎝ transmission _ data _ rate slot _ size ⎠ TABLE III. THE IEEE 802.11B MAC AND THE NETWORK -. total_backoff_period PARAMETERS THAT ARE USED IN THE SIMULATION = idle_slot_count + busy_slot_count Section Value -. a0(N,n)= idle_slot_count Data rate 11 Mbits/s Obtain the estimated number of active nodes Slot_time 20 μs log(a0 ( N, n)) −log(total _ backoff _ period ) SIFS 10 μsnest = DIFS 50 μs log(total _ backoff _ period −1) −log(total _ backoff _ period ) CWmin 31Step 2: Deciding the optimal CW CWmax 1023 Packet size exponential(1024) bytes Obtain the optimal CW Packet inter-arrival time exponential(0.1) sec -. CWoptimal= nest
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE CCNC 2010 proceedings 6 x 10 A. The network throughput 5 Fig. 3 indicates the throughput according to various backoff BEB EIED algorithms in the IEEE 802.11 WLAN. The efficiency standard 4.5 EILD of the DCF performs worse (as expected) when more stations PCB contend for the channel. Although the EIED algorithm takes an EBA 4 exponential decrease in the CW policy instead of resetting to CWmin when there is a successful transmission, the curve Throughput(bits/sec) decreases when there are more active stations in the system. 3.5 This means that the stations that are applying the EIED and the DCF algorithms make decisions with an unclear system status 3 and they quickly adjust the CW from the result of a single transmission. In contrast to the PCB, the EIED and the DCF, the throughput of the EILD and the EBA algorithms remains 2.5 high with respect to various system loads. These improvements mean that the stations that are using both the EILD and the 2 EBA algorithms adjust the CW value appropriately according 30 35 40 45 50 Number of nodes 55 60 65 70 to the load variation within the network. In the cases of both light and heavy loads, the EBA successfully determined the optimal backoff slot because the traffic measurement is Figure 3. The throughput vs. the number of nodes accurate. Overall, the EBA algorithm obtains high efficiency when it is compared with the other backoff algorithms in 4.5 various network conditions. 4 BEB EIED B. The average access delay 3.5 EILD The variation of the end to end packet delay according to Average access delay(sec) 3 PCB the number of active nodes is presented in Fig. 4. As expected, EBA the delay increases as the number of nodes increases. The 2.5 objective of the EBA algorithm is estimating the actual network status and setting the corresponding optimal CW to 2 precisely minimize the overheads in the system. In Fig. 4, the 1.5 EBA shows the advantage of overhead reduction and the EBA obtains the lowest delay among these backoff algorithms. The 1 delay of the EBA is around 50% less than that of a standard 0.5 DCF when n = 70. 0 30 35 40 45 50 55 60 65 70 C. The fairness Number of nodes Fairness among stations is an important problem in the BEB Figure 4. The average access delay vs. the number of nodes study, and it has been discussed by many research projects. The Fairness index can show if a resource is fairly allocated to each station. We use Jain’s fairness index formula. Jain’s fairness index 0.6 is calculated as ⎛ n ⎟2⎞ 0.5 ⎜ y⎟ ⎜∑ i ⎟ ⎜ i=1 ⎟ ⎝ ⎠ Fairness Index g ( y1 , y2 ,..., yn ) = n . (16) n ⋅ ∑ yi 2 0.4 i =1 0.3 BEB PCB Jain’s fairness index always lies between 0 and 1. A EIED fairness index of 1 indicates a throughput-fair algorithm . In 0.2 EILD Fig. 5, we present the fairness index of each backoff algorithm EBA among the stations. By using the simulation setup that was 0.1 30 35 40 45 50 55 60 65 70 described in the previous section, we executed the simulation Number of nodes for 10 iterations, and we calculated the average of the results. From Fig. 5, the proposed EBA algorithm has the most stability Figure 5. The Fairness index vs. the number of nodes when it is compared with the other contention algorithms. We also observe that the fairness index of the BEB, the EILD, the EIED and the PCB are both low and oscillatory. This
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE CCNC 2010 proceedingsphenomenon means that some stations occupy more channelcapacity than do other stations due to the differentunderstanding of the system status among stations. V. CONCLUSION The proposed EBA algorithm estimates the system statusby using the idle slot counts for the backoff duration, and itdetermines a proper contention window size that accuratelymatches the current network conditions. We compared theperformance of the proposed EBA with that of the conventionalalgorithms such as the IEEE 802.11 the DCF, the EIED, theEILD and the PCB. Our simulation results show that the EBAoutperforms the previously proposed algorithms for variousperformance metrics, and that the EBA dynamically adapts tothe variations of the amount of data traffics in the network. Based on the simulation results, we can use the proposedalgorithm in the future transportation information systemnamed as Telematics. The Telematics is a system where theinformation such as traffic jam, living, and emergency rescue,and etc. is exchanged between the vehicles. The Telematicsneeds more efficiency backoff algorithm because the variationof data traffics may be large due to the many vehicles’existence in the heart of city. Therefore the proposed EBA mayimprove the performance of Telematics system. In the future, we plan to explore how to implement ouralgorithm in the Telematics system. ACKNOWLEDGMENT"This research was supported by the MKE(The Ministry ofKnowledge Economy), Korea, under the ITRC(InformationTechnology Research Center) support program supervised bythe NIPA(National IT Industry Promotion Agency" (NIPA-2009-C1090-0902-0003) REFERENCES IEEE 802 Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications, IEEE Std., 1999. N. Song, B. Kwak, J. Song and M.E. Miller, “Enhancement of IEEE 802.11 distributed coordination function with exponential increase exponential decrease backoff algorithm,” in Proc. VTC 2003, Vol. 4, pp. 2775 – 2778, Orlando, USA, 22-25 April 2003. V. Bharghavan, A. Demers, S. Shenker, and L. Zhang, “MACAW: A Media Access Protocol for Wireless LAN’s,” in Proc. SIGCOMM’94, pp. 212-225, London, England, 1994. H. Liang, S. Zeadally, N. K. Chilamkurti and C. Shieh, “A Novel Pause Count Backoff Algorithm for Channel access in IEEE 802.11 based Wireless LANs,” in Proc. CSA 08 International Symposium, pp. 163 – 168, Hobart, Australia, 13-15 Oct. 2008. J. Lee, W. Kim and H. Kim. “Estimation of Number of Tags in ALOHA-based RFID Systems,” KICS, ’07-7 Vol. 32, No.7, 2007. J. R. Cha and J. H. Kim, "Dynamic Framed Slotted ALOHA Algorithm using Fast Tag Estimation method for RFID System," in Proc. CCNC2006, Las Vegas, USA, 8-10, Jan. 2006. Normal Lloyd Johnson and Samuel Kotz, Urn Models and Their Applications, Wiley, 1977. Q. Nasir and M. Albalt, “History Based Adaptive Backoff (HBAB) IEEE 802.11 MAC Protocol”, in Proc. CNSRC’08, pp. 533 – 538, Nova Sotia, Canada, 5-8, May. 2008. R. Jain, The art of computer systems performance analysis: techniques for Experimental Design, Measurement, Simulation, and Modeling, Wiley, New York, 1991.