Cross-Layer Performance Analysis of CSMA/iCA Based Wireless Local Area Network

Transcription

1 Wireless Pers Commun DOI /s Cross-Layer Performance Analysis of CSMA/iCA Based Wireless Local Area Network Subodh Pudasaini Seokjoo Shin Springer Science+Business Media, LLC Abstract Carrier sense multiple access with improvised collision avoidance (CSMA/iCA) has recently been proposed as an enhancement to CSMA/CA. It has been reported to be superior than the legacy counterpart in terms of throughput efficiency, packet transmission delay and quantitative fairness index. Nevertheless, the superiority has been shown assuming ideal network conditions: error-free physical layer (L1) and saturated (always non empty) queue at medium access control layer (L2). These strict assumptions, however, do not accurately hold in the real-world Wireless Local Area Networks since the wireless medium is generally errorprone and the arrival of the packets at L2 queue is generally bursty resulting in non-saturated queue occupancy. Thus, the reported performance, especially throughput, in such typical L1/L2 settings is not complete to understand the performance benefit that CSMA/iCA offers under the realistic network settings. In this paper, we relax the aforesaid ideal assumptions and present a cross-layer (L1/L2) performance analysis. Our cross-layer analytical model considers the effect of Rayleigh fading induced bit errors in L1 and non-saturated queue occupancy due to Poisson packet arrival at L2 queue. By virtue of the validated numerical results, we show that the CSMA/iCA consistently retains its throughput gain over CSMA/CA for the non-ideal wireless settings as well. Keywords CSMA/iCA Cross-layer analysis Rayleigh fading Unsaturated queue Markov process DCF WLAN 1 Introduction Although Wireless Local Area Network (WLAN) was originally conceived for the mere replacement of wired-lan, it is now one of the most preferred access technologies at homes, S. Pudasaini S. Shin (B) Department of Computer Engineering, Chosun University, Gwangju , Korea sjshin@chosun.ac.kr S. Pudasaini subodh@chosun.kr

2 S. Pudasaini, S. Shin enterprises, and public hot-spots. This paradigm shift happens due to the decreasing cost of WLAN networking equipments like Access Points (APs) and Network Interface Cards (NIC); the gradual increment in WLAN data rates; and the growing use of laptops and Personal Digital Assistants (PDAs). According to IEEE [1], the de facto standard that covers the lower two layers of WLAN protocol stack, networking between multiple wireless stations (STAs) can be made in two different modes: infrastructure mode and ad hoc mode. In the former, STAs communicate with each other and with the wired network via an intermediate central device called AP while in the later STAs communicate directly with each other without involving any intermediate AP. For both operating modes, Medium Access Control (MAC) protocol has been specified. Distributed Coordination Function (DCF) is the base MAC protocol for both operating modes while Point Coordination Function (PCF) can optionally be used in the infrastructure mode. PCF, however, is barely implemented in commercial AP and NIC cards. Distributed Coordination Function (DCF) is a contention-based MAC protocol. It defines both signalling and timing functionalities of CSMA/CA-based access mechanism. It works fairly well for light traffic load. As the number of contending STAs increases, however, severe contention and collisions degrade WLAN performance. Not only the network throughput is decreased, but also individual packets are delayed and the throughput-fairness of contending STAs is reduced. To concurrently address these problems, Carrier sense multiple access with improvised collision avoidance (CSMA/iCA), a simple enhancement to CSMA/CA, has recently been presented in [2]. CSMA/iCA has been reported to be superior than the legacy counterpart in terms of throughput efficiency, packet transmission delay and quantitative fairness index. In [2], a tridimensional (3D) Discrete Time Markov Chain (DTMC) was developed to characterize the backoff process of CSMA/iCA. Based on the steady-state solution of the DTMC, adopting a renewal reward process based network model, throughput efficiency was analytically estimated. To simplify the analysis, following assumptions were made: Physical layer (L1) is error-free and MAC queue (L2) is saturated (always non empty). These assumptions, however, do not accurately hold in the real-world WLANs since the wireless medium is generally error-prone and the arrival of the packets at L2 queue is generally bursty resulting in non-saturated queue occupancy. Thus, the estimated throughput in [2], in such typical L1/L2 settings, is not complete to understand the actual performance benefit that CSMA/iCA can offer under the realistic network settings. In this paper, we relax those ideal L1 and L2 assumptions and present a cross-layer (L1/L2) performance analysis of CSMA/iCA based WLAN. The new cross-layer analytical model considers the effect of Rayleigh fading induced bit errors in L1 and non-saturated queue occupancy due to Poisson packet arrival at L2. The new cross-layer analysis results a general analytical model which reproduces the model in [2] as its special case when the packet error probability due to channel noise is zero and the probability of packet availability at L2 queue is 1. Note that this paper is an extension of our previous paper [3]. The extension includes the validation of the presented analytical model using ns-2. The rest of the paper is organized as follows: In Sect. 2,CSMA/iCA is revisited. The new analytical throughput estimation model for CSMA/iCA based WLAN is presented in Sect. 3. The performance evaluation results obtained from the new analytical model along with the ns-2 simulation based model validation results are presented in Sect. 4. Finally, in Sect. 5, we conclude the paper.

3 Cross-Layer Performance Analysis 2 Quick Overview of CSMA/iCA Carrier sense multiple access with improvised collision avoidance (CSMA/iCA) is an enhanced CSMA/CA with an improvised collision mitigation feature [2]. Due to space limitation, we briefly review CSMA/iCA with only necessary information that is required to understand the upcoming cross-layer analysis. Carrier sense multiple access with improvised collision avoidance (CSMA/iCA) can be understood as an enhancement to the legacy counterpart pertaining to CA mechanism. In the legacy CA mechanism, the Transmission Probability (TP) of each contending STA is temporally adjusted by changing only the size of Contention Window (CW) upon witnessing success or collision of the transmitted packet. On the other hand, in CSMA/iCA TP is adjusted both by regulating CW and Contention Slot Selection Distribution (CSSD) over CW. In other words, in CSMA/iCA a dynamic non-uniform CSSD is specified over CW, unlike the statically flat (uniform) distribution in the legacy CA mechanism. The shape of the CSSD is regulated using some mapping relations based on locally-available History Backoff Value (HBV). We denote the distribution as g(k j), wherek is any of the slot in CW and j is HBV. Updating the CSSD based on HBV contributes in mitigating the collisions. The collision mitigation achieved with CSMA/iCA can be observed either from local perspective of the individual contending STAs or from global perspective of the broadcast channel. From the local perspective, the collision mitigation mechanism looks simple. In consists of two straightforward processes: classification and prioritization of the contention slots over CW. Instead of blindly selecting any slot in CW for making access attempt, the proposed method first uses HBV to classify the contention slots in two sets: Relatively High Collision Prone (RHCP) and Relatively Less Collision Prone (RLCP), and prioritizes the RLCP slots by adaptively tuning the shape of the CSSD according to HBV. From the global perspective, a pool of backlogged access attempts from multiple STAs is logically divided into numerous smaller groups with probabilistically separated smaller CWs so that both the intra-group and inter-group access collision chances would reduce. For more details about the global and local perception, and operational mechanism of CSMA/iCA we refer readers to [2]. 3 Cross Layer Analysis of CSMA/iCA Figure 1 depicts the high-level framework of the cross-layer analytical model. It consists of two sub-models: a user model and a network model. As to be described in the subsequent sections, the user model which is based on DTMC takes into account the channel arbitration process of CSMA/iCA based WLAN considering the influence of bit errors on L1 and Packet Arrival Rate (PAR) at L2 queue. The output of the user model along with network population size and the timing details of the signalling mechanisms of IEEE DCF are used as inputs to the Renewal Reward Process (RRP) based network model to estimate the performance of interest. 3.1 System Model We consider a WLAN with N contending STAs where their channel access opportunities are arbitrated according to a new DCF. DCF in which CSMA/CA is replaced with CSMA/iCA. We consider a M/M/1/K queue at L2 of each STA. Packets arrive at the queue in a Poisson manner with exponentially distributed inter-packet arrival time with mean rate λ a

4 S. Pudasaini, S. Shin Fig. 1 High-level framework for the considered analytical model and are serviced at the rate λ s 1. Hence, the probability q that the queue remains non empty can be expressed as 1 λ a /λ s q = 1. (1) 1 (λ a /λ s ) K +1 Regarding L1, we consider it to be noisy; the transmitted packets might possibly experience bit errors. Note that probability of bit errors primarily depends on the utilized modulation technique and the channel characteristics. In this current work, we have considered Differential Binary Shift Keying (DBPSK) and a typical Rayleigh faded channel. For DBPSK, probability of bit error, p b, over Additive White Gaussian Noise (AWGN) channel is [4] p b (γ ) = 1 exp( γ), (2) 2 where γ is the received instantaneous Signal to Noise Ratio (SNR). In Rayleigh faded channel, γ is a random variable with the following probability distribution function f (γ ) = 1 ( ) γ exp, (3) γ 0 where γ 0 is the average SNR. Thus, bit error probability for DBPSK-modulated Rayleigh faded channel can be estimated as p b = = 0 0 p b (γ ) f (γ )dγ γ exp( γ) 1 γ 0 exp ( γ γ 0 ). (4) Note that (4) reflects the bit error probability of a IEEE b STA operating in its lowest available channel rate (1 Mbps). For the other available channel rate options, 5.5 and 11 Mbps, p b can be calculated as elaborated in [5]. Based on p b in (4), packet error probability in L2 can be calculated. To calculate the packet error probability in L2, we assume that the bit errors are identically and independently distributed (i.i.d) over the whole packet. Due to such assumption of i.i.d bit errors, packet error probability for a packet x can be expressed as p x e = 1 (1 p b) L x, (5) where x is a type of the packet (either DATA or ACK) and L x is the length of x in number of bits. It is noteworthy to mention that packet error probability should take into account 1 For a given WLAN configuration, λ s can be calculated as discussed in the last paragraph of Sect. 3.

5 Cross-Layer Performance Analysis both packet error probabilities of DATA and the subsequent ACK (recall the two way DATA- ACK exchange in the basic access mechanism of DCF). Assuming independence between bit errors of DATA and ACK packets, effective packet error probability observed at L2 can be written as p e = p data e + p ack e pe data pe ack. (6) Note that packet transmission fails not only due to channel errors, but also due to collisions. Collision happens with the following probability when more that one STAs transmit at the same time p c = 1 (1 τ) N 1, (7) where τ is the transmission probability of each contending STA. Finally, assuming independence between the chances of packet collision and packet error, the overall transmission failure probability for the considered L1/L2 settings can be expressed as p eq = p c + p e p c p e 3.2 Markov Chain for CSMA/iCA = 1 (1 τ) N 1 (1 p e ). (8) In this section we present a 3D DTMC that we develop to emulate the backoff procedure of the CSMA/iCA. The developed DTMC is based on the seminal 2D DTMC in [6]. However, there are following differences: (1) An additional dimension is added to track every possible HBVs, (2) Channel errors are accounted as in [7], (3) L2 queue does not necessarily always has packet to transmit. Figure 2 depicts the 3D DTMC. At any time t, a contending STA can be in any of the states (rounded rectangles in the figure) (i, j, k) or Idlewhere i [0, m] (m is maximum allowed retransmission limit) is the contention stage, j is the current value of the backoff counter, and k is the backoff value with which the contention was initiated in the stage i.notethat,in CSMA/iCA, contention window W i for stage i is regulated as per binary exponential backoff policy but there is subtle difference in CSSD over W i as previously described in Sect. 2. In the 3D DTMC, two types of transitions can happen: intra-contention stage transition and inter-contention stage transition. Intra-contention stage transition happens within any stage i when an idle slot is detected and j = 0. If j is zero and an idle slot is detected, a STA transmits its packet. Based on the result of transmission attempt, success or failure, inter-contention stage transition takes place. In the figure, both intra and inter-contention stage transitions are marked with a line with a filled arrow head. By adopting the conventional notation {i 1, j 1, k 1 i 0, j 0, k 0 } to denote the transition from (i 0, j 0, k 0 ) to (i 1, j 1, k 1 ) with probability P{i 1, j 1, k 1 i 0, j 0, k 0 }, all the possible one-step transitions in the 3D DTMC and their corresponding probabilities can be written as follows: 1. P{i, j, k i, j + 1, k} =1; i [0, m], j [0, W i 2], k [0, W i 1], 2. P{0, j, k i, 0, k} = q(1 p eq) ; i [0, m 1], k [0, W i 1], k [0, W i 1], j = k, 3. P{0, j, k m, 0, k} = qp eq + q(1 p eq) ; k [0, 1], j = k, k [0, W m 1],

6 S. Pudasaini, S. Shin Fig. 2 Emulation of the backoff procedure of CSMA/iCA with a 3D DTMC considering L1 channel errors and unsaturated L2 queue 4. P{i, j, k i 1, 0, k} =p eq g i (k k); i [1, m], j [0, W i 1], k = j, k [0, W i 1 1], 5. P{I i, 0, k} =(1 q)(1 p eq ); i [0, m 1], k [0, W i 1], 6. P{I m, 0, k} =1 q; k [0, W m 1],

7 Cross-Layer Performance Analysis 7. P{0, k, k I }= p a ; k [0, W i 1 1], k = k, 8. P{I I }=1 p a. The above expressions account, respectively, for: 1. Probable transitions when an idle slot is detected in stage i; contention stage i remains the same, backoff counter j is decremented by 1, and the history of the originally selected backoff value k is copied. 2. Probable transitions when L2 queue is found to be non empty upon making a successful packet transmission during contention stage i ; contention stage i isresetto0, CWis reset to, and slot k is selected randomly (uniformly) over. 3. Probable transitions when L2 queue is found to be non empty after making either a successful or a failed packet transmission in contention stage m; contention stage m is resetto0,cwisresetto, and slot k is selected randomly (uniformly) over. 4. Probable transitions after making an unsuccessful packet transmission in contention stage i 1; contention stage i 1 is increased by 1, CW is doubled, and slot k is selected randomly following g i (k k). 5. Probable transitions when L2 queue is empty after a successful packet transmission in contention stage i; STA enters into Idle state and waits for new packet to arrive. 6. Probable transitions when L2 queue is empty after either a successful or a failed packet transmission in contention stage m; STA enters into Idle state and waits for new packet to arrive. 7. Probable transitions when a new packet arrives in the empty L2 queue; backoff procedure is invoked with the initial contention window.notethatp a = 1 e λp I T I where definition of P I and T I areavailablein(19)and(20). 8. Probable transitions when no new packets arrive in the empty L2 queue. 3.3 Analysis of Markovian Process of CSMA/iCA In the 3D DTMC, let the stationary distribution of a tagged STA to be in state (i, j, k) and state Idle be b i, j,k and b I, respectively. Furthermore, let (i, j, X) be a compound state representing all the states having k j for the given i and j,andb i, j,x be the probability of being in the compound state (i, j, X). Mathematically, B i, j,x is equal to (b i, j, j bi, j, j+1 b i, j,wi 1). In Fig. 2, note that the inter-contention stage transitions can occur only from the compound state B i,0,x i [0, m]. Hence, the probability of being in compound state (i, 0, X) can be expressed as a function of the probability being in state (i 1, 0, X), i.e., B i,0,x = p eq B i 1,0,X, 0 < i m. (9) Since the 3D DTMC is regular, B i, j,x for 0 < i m can be written as W i 1 B i, j,x = p eq B i 1,0,X = p i eq B 0,0,X y= j W i 1 y= j W i 1 1 z=0 W i 1 1 z=0 g i (y z) g i (y z). (10)

8 S. Pudasaini, S. Shin Similarly, for the case i = 0, B 0, j,x can be written as [ ] B 0, j,x = W m 1 0 j qb m,0,x + q(1 p eq ) B l,0,x + p a b I. (11) Likewise b I can be written as m 1 b I = (1 q)(1 p eq ) B l,0,x + (1 q)b m,0,x + (1 p a )b I, (12) which can be further simplified to l=0 l=0 p a b I = (1 q)(1 p eq ) 1 pm eq (1 p eq ) + (1 q)pm eq B 0,0,X b I = 1 q B 0,0,X. (13) p a Using (13), (11) can be simplified to B 0, j,x = j [ m 1 q(1 p eq ) l=0 ] peq l B 0,0,X + qb m,0,x + p a(1 q) B 0,0,X p a = j B 0,0,X. (14) Equations (10) and(14) express B i, j,x for 0 i m and Eq. (12) expresses b I as a function of B 0,0,X. Hence, B 0,0,X can be determined by imposing the following normalization condition: 1 = b I + m i=0 W i 1 j=0 B i, j,x = 1 q W 0 1 W j i 1 B 0,0,X + B 0,0,X + p a W j=0 0 i=1 j=0 Upon simplifying (15), B 0,0,X = 1 q + p a W 0 1 j=0 j + m i=1 m W i 1 j=0 W i 1 1 z=0 W i 1 1 B i,0,x W i 1 y= j z=0 W i 1 y= j peq i g i (y z) g i (y z). (15) 1. (16) Since packet transmission can happen from any of the compound state B i,0,x, the probability that the tagged STA transmits in a randomly chosen slot can be expressed as ( ) m 1 peq m+1 τ = B i,0,x = B 0,0,X. (17) 1 p eq i=0 By solving a nonlinear system formed by pair of equations in (17)and(8), two unknowns τ and p eq can be obtained. With the known τ, the WLAN throughput can be estimated using the RRP based network model, to be described next.

9 Cross-Layer Performance Analysis 3.4 Network Model for Throughput Estimation We adopt the RRP based network model in [6] to estimate the long-run average throughput ξ. Let us consider that transmission attempts from N contending STAs, each of which transmits packet with probability τ derived in the previous subsection, are independent events and they repeat over time. The outcome space for such events is {I, S, C, E d, E a }, where the notations have the following interpretations, I: no access attempt (channel is idle); S: attempt is successful and the reception is error free as well; C: attempt ended in a collision; E d : attempt is successful but the reception is not error free; and E a : attempt is successful and the reception is error free but the returned acknowledgement is suffered from channel error. Considering the outcome space for such repetitive events, inter event duration T can be estimated as follows: E[T ]= P x T x, (18) x (I,S,C,E d,e a ) where P x is the probability that event x happens and T x is the duration for which the event x lasts. With the assumption that all N contending STAs can hear each others transmission (i.e connected single-hop network), P x can be obtained as follows: P I = (1 τ) N, ( )( P S = Nτ(1 τ) N 1 1 pe data P C = 1 (1 τ) N Nτ(1 τ) (N 1), P Ed = Nτ(1 τ) N 1 pe data (, ) P Ea = Nτ(1 τ) N 1 1 pe data 1 p ack e ), p ack e. (19) T x for the basic access mechanism can be obtained by considering the signaling mechanisms specified for DCF in [1] as follows: T I = Physical Slot Duration, T S = DIFS + T H + T D + SIFS + T A, T C = T H + T D + EIFS, T Ed = T C, and T Ea = T S, (20) where time required to transmit H bits of header, L D bits of data (D), and L A bits of acknowledgement (A) are T H = H MAC R D + H PHY R C, T D = L D R D, and T A = L A+H PHY R C, respectively, for the given data rate R D and control rate R C bits/s. It should be noted that EIFS = SIFS + T A + DIFS.ForthegivenN, nonlinear system formed by (8) and(17) can be solved numerically to get the values of two unknowns τ and p eq. Based on these values, P x x (I, S, C, E d, E a ) in (19) can be calculated. T x in (20) can be obtained for the given L D. Upon obtaining both P x and T x, E[T ] in (18) can be obtained. Once E[T ] is estimated, ξ can be calculated by considering the reward R per renewal event T as follow: where E[R]=P S L D. ξ = E[R] E[T ], (21)

10 S. Pudasaini, S. Shin Note that the presented 3D DTMC can be utilized to estimate ξ of the saturated WLAN as well. For this, q in (16) should be assigned 1. Let the solution of non-linear system in (8) and (17) for saturated WLAN be τ s and corresponding E[T ] in (18)beE[T s ]. For the known τ s and E[T s ], λ s in (1) can be obtained as follows: λ s = τ s(1 τ s ) N 1. (22) E[T s ] 4 Performance Results 4.1 Model Validation We validate the accuracy of the theoretical results obtained via the proposed analytical model by comparing them to the simulation results obtained from ns-2 [8]. For doing that we not only implemented the CSMA/iCA for IEEE DCF in ns-2 (version 2.29) but also modified the signal reception model because the conventional PHY layer implementation does not take into account the effect of bit error on the transmitted packets. In the modified signal reception model, bit error on the transmitted packets are accounted (when determining the success or failure of a received signal) as described in [9], along with the three conventional SNR based thresholds: carrier sense threshold (CSThresh), receive threshold (RxThresh), and capture thresh (CPThresh). A typical infrastructure Basic Service Set (BSS) is considered where the access point is located in the center of a circular network area of radius 20 m. STAs are randomly located over the circumference of the network area. All STAs transmit 1,024 Bytes UDP packet to the access point. The inter-packet arrival time at each STA is exponentially distributed with mean rate of λ a packets/s. No Ad-Hoc Routing Agent (NOAH)[10] is used to bypass the effect of routing in the network s performance. The considered MAC parameters are summarized in Table 1. Regarding the propagation model, the shadowing model is adopted. The path loss exponent is considered to be 3.6. CSThresh and RxThresh are considered to be 94 and 95 dbm, respectively, while the transmission power is considered to be w. It is worth mentioning that we have fixed the value of CPThresh to relatively very higher value than its default value of 10 db because in our simulation we do not want to capture any packets in the situations when simultaneous transmissions happen, as our analytical model does not consider the case related to capture effect. Under the aforesaid simulation settings, we observe the throughput for the the two different scenarios: (1) varying number of saturated STAs (whose queue has always packet to transmit) under a given bit error rate and (2) varying packet arrival rate for a given number of STAs. Each simulation experiment for a particular scenario was run for 100 s after a 10 s initialization. All of the presented throughput values are the average of the 10 values obtained from independently repeated simulation experiments. Simulation results along with the results from the analytical model are shown in Fig. 3. In particular, Fig. 3a depicts the throughput when there were varying number of saturated STAs (5 30) for a special case when bit error rate over the channel was Figure 3b depicts the throughput for the case of varying packet arrival rates (10 35 packets/s) when there were 10 STAs and the bit error rate was Simulation results for these two scenarios adequately match with the corresponding theoretical results, though the throughput is slightly over estimated (<1 %) when either the number of stations STAs or packet arrival rate was high.

11 Cross-Layer Performance Analysis Table 1 Considered network and channel parameters for numeric analysis and simulation Parameters Values Slot time 20 μs DIFS 50 μs SIFS 10 μs MAC header 224 bits PHY header 192 bits ACK packet 112 bits + PHY header R D and R C 1 Mbps Minimum CW 32 Maximum CW 1, 024 Retransmission limit 6 Channel model DBPSK modulated Rayleigh-faded channel Queue length (K) 50 Packet arrival rate Variable (0 50 packets/s) Packet length 512, 1, 024Bytes Number of STAs (N) Variable (up to 70) (a) (b) Fig. 3 Theoretical and simulated throughput of the CSMA/iCA based WLAN: a varying number of station, b varying packet arrival rate 4.2 Throughput Comparison of CSMA/CA and CSMA/iCA Once the accuracy of the analytical throughput estimation model for CSMA/iCA based WLAN had been validated, it was used to generalize the results for different network and channel conditions. The results are compared with the corresponding results of the conventional CSMA/CA based WLAN. For both WLANs, we firstly analyze the effect of Rayleigh fading induced L1 bit errors on the throughput considering a special scenario in which each

12 S. Pudasaini, S. Shin (a) (b) Fig. 4 Throughput of a saturated WLAN with two different MAC protocols, DCF with CSMA/CA and DCF with CSMA/iCA, under channel errors: a fixed channel error with increasing traffic load, b varying channel error for a fixed traffic load contending STA has saturated L2 queue, and then we generalize that result for the case when the contending STAs do not necessarily have saturated L2 queue. It is a well established fact that channel errors degrade the performance of any wireless networks. Figure 4 depicts the trend how channel errors degrade throughput of CSMA/CA

13 Cross-Layer Performance Analysis Fig. 5 Throughput of a WLAN with two different MAC protocols, DCF with CSMA/CA and DCF with CSMA/iCA, under different packet arrival rates and CSMA/iCA based saturated WLANs. As can be noted in Fig. 4a, for a given number of contending STAs with fixed-size homogeneous packets (1,024 Bytes), throughput of both WLANs decreases as BER increases (correspondingly, SNR decreases). Since channel error not only depends on the SNR but also on the size of the transmitted packets, Fig. 4b further elaborates Fig. 4a by including additional case studies for different packet sizes. Figure 5 compares the throughput of CSMA/iCA and CSMA/CA based WLANs over an error-prone channel (average SNR of 37 db) for the broad range of packet arrival rates. In both WLANs, for the given number of contending STAs, as the packet arrival rate at each contending STAs increases, the network throughput linearly increases up to their respective maximum achievable throughput limits and settles as a plateau thereafter. Lets denote the arrival rate that resulted the plateau be critical arrival rate. For the arrival rates less than the critical arrival rate, CSMA/iCA based WLAN offers similar throughput as CSMA/CA based WLAN does. Interestingly, CSMA/iCA based WLAN offers higher throughput for the arrival rates greater than or equal to critical arrival rate. This result implies that as the network traffic load increases, either due to the increase in the number of contending STAs or due to the high packet arrival rate at each contending STAs, CSMA/iCA based WLAN performs better. 5Conclusion In this paper, we have presented an analytical model to estimate throughput of a CSMA/iCA based WLAN considering cross-layer (L1/L2) details. The presented analytical model not only takes into account the improvised collision avoidance feature of CSMA/iCA but also the effects of channel errors and packet arrival rate at contending STAs. Through numer-

14 S. Pudasaini, S. Shin ical results, we have shown that CSMA/iCA consistently offers throughput benefit over CSMA/CA in every possible channel and traffic conditions. Simulation results adequately confirm the validity of the presented analytical model. Acknowledgments This study was supported by research fund from Chosun University, References 1. IEEE Std (2007). Part 11: Wireless LAN medium access control (MAC) and physical layer (PHY) specifications. 2. Pudasaini, S., Shin, S., & Kim, K. (2012). Carrier sense multiple access with improvised collision avoidance and short term fairness. Wireless Networks. doi: /s Pudasaini, S., & Shin, S. (2011). Cross layer analysis of CSMA/iCA based wireless local area network. In third international conference on ubiquitous and future networks (pp ). 4. Proakis, J. G. (2000). Digital communications (4th ed.). New York: McGraw Hill. 5. Mahasukhon, P., Hempe, M., Sharif, H., & Chen, H. (2007). BER analysis of b networks under mobility. In IEEE international conference on communications (pp ). 6. Bianchi, G. (2000). Performance analysis of the IEEE distributed coordination function. IEEE Journal on Selected Areas in Communications, 18(3), Ni, Q., Li, T., Turletti, T., & Xiao, Y. (2005). Saturation throughput analysis of error-prone wireless networks. Wireless Communications and Mobile Computing, 5(8), Network Simulator Daneshgaran, F., Laddomada, M., Mesiti, F., Mondin, M., & Zonolo, M. (2008). Saturation throughput analysis of IEEE in the presence of non ideal transmission channel and capture effects. IEEE Transaction on Communications, 56(7), No adhoc routing agent (NOAH). Author Biographies Subodh Pudasaini received the B.E. degree in electrical and electronics engineering from Kathmandu University, Nepal, in 2005 and the M.E degree in computer engineering from Chosun University, Korea, in Currently, he is a Ph.D. candidate at department of computer engineering, Chosun University, majoring in wireless communications and networks. His current research interests include design and analysis of wireless medium access control protocols and QoS provisioning in various distributed networks including local, personal, and body area networks.

15 Cross-Layer Performance Analysis Seokjoo Shin received the B.Eng. in avionics engineering from Korea Aerospace University, Korea in 1997, and the M.Eng. and Ph.D. degrees in information and communications engineering from Gwangju Institute of Science and Technology (GIST), Korea, in 1999 and He joined the Mobile Telecommunication Research Laboratory, Electronics and Telecommunications Research Institute (ETRI), Korea, in In 2003, he joined Chosun University, where he is currently associate professor in the Department of Computer Engineering. His research interests include the resource management and access control protocols for wireless communication systems.