Performance Analysis of IEEE MAC Protocol with Different ACK Polices

Transcription

1 Performance Analysis of IEEE MAC Protocol with Different Polices S. Mehta and K.S. Kwak Wireless Communications Research Center, Inha University, Korea Abstract. he wireless personal area network (WPAN is an emerging wireless technology for future short range indoor and outdoor communication applications. he IEEE medium access control (MAC is proposed, specially, for short range high data rates applications, to coordinate the access to the wireless medium among the competing devices. his paper uses analytical model to study the performance analysis of WPAN (IEEE MAC in terms of throughput, efficient bandwidth utilization, and delay with various acknowledgment schemes under different parameters. Also, some important observations are obtained, which can be very useful to the protocol architectures. Finally, we come up with some important research issues to further investigate the possible improvements in the WPAN MAC. Keywords: Analytical Modeling, Performance Analysis, MAC Protocol, IEEE Introduction he IEEE standard MAC layer [1] is based on a centralized, connection oriented topology which divides a large network into several smaller ones termed piconets. A piconet consists of a Piconet Network Controller (PNC and DEVs (DEVices. he DEV is designed to be low power and low cost. One DEV is required to perform the role of PNC (Piconet Coordinator, which provides the basic timing for the piconet as well as other piconet management functions, such as power management, Quality of Service (QoS scheduling, security, and so on. he standard also allows for the formation of child piconets and neighbor piconets. he original piconet is called the parent piconet and the child/neighbor piconets are called the dependent piconets. hese piconets differ in the way they associate themselves to the parent piconet. IEEE standard supports multiple power saving modes and multiple acknowledgement ( policies (NO, Imm-, and Dly-,. It is very robust and supports coexistence with the other WLAN technologies such as IEEE In IEEE MAC protocol, although communications are connection based under the control of the PNC, connections and data transfer can be made with peer to peer connections. In IEEE MAC protocol, the channel time is divided into superframes, which each superframe beginning with a beacon. he superframe is composed of the three major parts: the beacon, the optional

2 contention access period (CAP, and the channel time allocation period (P or channel allocation time (. Wireless channel is usually vulnerable to errors. Error control mechanisms should be designed in MAC protocol to provide a cretin level of reliability for the higher network layers. In accordance with that, IEEE defines three types of acknowledgment mechanisms for s: the No-, Imm-, and Dly- mechanisms. For the Imm- and Dly- mechanisms, retransmission is adopted to recover corrupted frames in previous transmissions. For the No- mechanism, is not sent after a reception. In the Imm- mechanism, each frame is individually ed following the reception of the frame. he Dly- mechanism allows the source to send multiple frames without waiting for individual s. Instead, the s of the individual frames are grouped into a single response frame to be sent to the source DEV. In [2] and some other literature proposed implied-acknowledgment (Imp- for bidirectional communication. Implied acknowledgement (Imp- permits a to be used bi-directionally within a limited scope. Implied policy is not allowed in the CAP to avoid ambiguities between a frame that is transmitted in response to a frame with an implied request and a frame that is transmitted independently when the original frame was missed. o reduce the overhead of the IEEE MAC, we combine the frame aggregation concept [3] and Dly- mechanism with different burst sizes, and we define this new mechanism as Dly--AGG. he idea of frame aggregation is to aggregate multiple MAC frames into a single (or approximately single transmission. All these policies have a large impact on the throughput, delay, and channel utilization of the network and required a detailed study to find overall performance of the network. In this paper, we present the performance analysis of IEEE from protocol architecture s point of view. Furthermore, we show the effect of aggregation with Dly-, i.e. Dly--AGG, on the network performance. 2 Related Works o the best of our knowledge, there is little work on the performance or channel analysis of IEEE MAC with respect to different policies, under different parameters. However, a large amount of literature is available on IEEE MAC scheduling, optimization of superframe size, various traffic analyses and so on. Some of the important related works are as follow. In [4] authors, presents the implementation of IEEE module in ns-2 and discuses various experimental scenario results including various scheduling techniques. Specially, to investigate the performance of real-time and best-effort traffic with various super frame lengths and different policies. In [5] authors, presents a novel sharing protocol, called VBR-M that enables the sharing of s belonging to streams with the same group identity. his feature allows the proposed protocol to exploit the statistical characteristics of variable bit rate (VBR streams by giving unused time units to a flow that requires peak rate allocation. And they presents two optimizations to VBR-M, namely VBR-Blind and VBR- okenbus, as well. In [6, 7], the authors proposed an adaptive Dly- scheme for

3 both CP and UDP traffic. he first one is to request the Dly- frame adaptively or change the burst size of Dly- according to the transmitter queue status. he second is a retransmission counter to enable the destination DEV to deliver the MAC data frames to upper layer timely and orderly. While later is more focused on optimization of channel capacity. Both papers laid a good foundation in simulation and analytical works of IEEE MAC protocol. Similarly, in [8] authors, formulate a throughput optimization problem under error channel condition and derive a closed form solution for the optimal throughput. he work presented in [8] is close to our work but their analysis scope is limited only in terms of throughput analysis. While our work span covers the delay, throughput, and channel utilization with different policies under frame aggregation and error channel condition. he paper is outlined as follows. In section 3 we present the performance analysis from protocol architecture s point of view and finally, conclusions and future work are drawn in section 4. 3 Performance Analysis of IEEE MAC Protocol In this section, we present the performance analysis of IEEE MAC to answer several questions like optimization of payload, optimization of policies, effect of aggregation etc., under various parameters. 3.1 Analytical Model We use ground works of Bianchis s model [9] and [8] to present our analysis work. We divide our analysis in two parts; analysis and CAP analysis. able 1 shows the notations that we used for the analytical model. Short Inter Frame Space (SIFS time SIFS Distributed Coordinate Function Inter Frame Space (DIFS DIFS time Minimum Inter Frame Space (MIFS time MIFS CW Minimum back-off window size min ransmission time of the physical preamble pre. ransmission time of the PHY header PHY L MAC overhead in bytes MAC H L size in bytes L Payload size in bytes Data ransmission time of MAC overhead MAC H transmission time ransmission time for the payload Data f he time for a transmission considered failed during CAP CAP he time for a transmission considered successful during CAP s CAP f he time for a transmission considered failed during CA he time for a transmission considered successful during CA s he time-out value waiting for an O h Normalized throughput during a CAP time CAP

4 h Normalized throughput during a time n Burst size in packets burst he throughput is given by able 1. Parameters notations ransmitted Data h = ransmission Cycle Duration (1 We assume a Gaussian wireless channel model. he channel bit error rate (BER, denoted as p e ( < p e < 1, can be calculated via previous frames or some other mechanism. How to obtain p e is way out of the scope of this paper. From [9], a frame with a length L in bits, the probability that the frame is successfully transmitted can be calculated as p s = (1 p L (2 Here, for the simplicity we assume that a data frame is considered to be successfully transmitted if both data frame and are successfully transmitted in different mechanism policies. We use Imm_, No_, D_, and D_AGG_ to denote the immediate acknowledgement, No acknowledgement, delay acknowledgement, and delay acknowledgement with aggregation, respectively. hen we can define p for different mechanisms as follows s p, Imm_=(1 p s ( LData + LMAC H + L Imm *8 e e ( L + L *8 p Data MAC H,No_=(1 p s p,dly_=(1 p s p, Dly_-AGG=(1 p s s e e ( L + L + L *8 Data MAC H Dly e ( L + L + L + L *8 Data MAC H MAC Hs Dly A successful transmission time during a is given by 1 = ( + MIFS Data MAC H pre PHY ( 2* + 2* SIFS Data MAC H pre PHY MAC PHY Imm ( ( + MIFS Data MAC H pre PHY SIFS MAC PHY Dly pre ( K * forno forimm K fordly Data * fordly AGG MAC H MAC HS pre PHY SIFS MAC PHY Dly AGG from (2, (3 and (4 the normalized throughput during a is given by (3 (4 1 Readers are advice to have a look at table 1 while referring equations as we couldn t explain every parameter due to space limitation.

5 h ps, No_ LData*8 for No s ps, Imm_ LData*8 for Imm s = ps, Dly_ KLData*8 for Dly s ps, Dly_ AGGKLData*8 for Dly AGG s o demonstrate the effect of n-dly- and n-dly--agg on bandwidth utilization, we define a metric named maximum effective bandwidth (MEB based on [1], which is a fraction of time the channel is used to successfully transmit data frames versus the total channel time. he maximum effective bandwidth utilization during a /CAP slot is given by (5 MEB LData ps, Dly_ nburst. fordly s = LData ps, Dly_ AGG nburst. for Dly AGG s (6 MEB CAP n 1 LData nψ(1 ψ ps,dly_, nburst. fordly scap = n 1 LData nψ(1 ψ ps, Dly_ AGG nburst. fordly AGG scap During the CAP, we use the analytical model similar with CSMA/CA mechanism of IEEE Also, we study how to optimize the channel throughput using different policies under error channel condition. When the Imm- mechanism is used in CAP, each node adopts CSMA/CA with binary exponential backoff. When NO- mechanism is used, each node will use a fixed backoff window as it has no knowledge whether or not the transmitted data frame is successfully received. If the Dly- mechanism is used, as long as the backoff timer of a node reaches zero, the node will first send a number (K of data frames each separated by an MIFS and a delay-request frame separated by an MIFS, and wait for the. If a burst transmission (of K data frames is considered successful, the sender will reset the backoff window to the initial value; otherwise, the backoff window will be doubled. Dly--AGG follows the same backoff procedure as Dly-. From [9], the failure probability of a transmission during a CAP is given by

6 1 (1 p ps, Imm_ forimm_ pf = 1 (1 p ps, Dly_, fordly_ 1 (1 p ps, Dly _ AGG, for Dly _ AGG For n number of stations, the probability of a transmitted frame collision is given by p ψ (7 1 = 1 (1 n (8 ψ, probability of a station to transmit during a generic slot time is also depends on number of retry limit. hen, the probability of the busy channel is given by pb = 1 (1 ψ n (9 From (8 and (9, the probability of a success transmission occurs in a slot time is given by n 1 nψ(1 ψ ps, No_, forno n 1 nψ(1 ψ ps, Imm _, forimm _ ps = n 1 n ψ(1 ψ ps, Dly_, for Dly_ n 1 nψ(1 ψ ps, Dly_ AGG, fordly_ AGG A successful transmission time during a CAP is given by scap = ( + MIFS Data MAC H pre PHY ( 2* + 2* SIFS Data MAC H pre PHY MAC PHY Imm ( ( + MIFS Data MAC H pre PHY SIFS MAC PHY Dly pre ( * 2* Data MAC H MAC HS pre PHY SIFS MAC PHY Dly AGG CW for No CW for Imm CW+ K for Dly AC K CW + K fordly AGG (1 (11 where CW represents the average back-off time. he average back-off defines the back-off duration for light loaded networks, i.e. when each station has access to the channel after the first back-off attempt and is given by CW. min CW = slot 2 A failure transmission time during a is given by (12

7 f = ( + MIFS Data MAC H pre PHY ( + SIFS Data MAC H pre PHY o ( ( + MIFS Data MAC H pre PHY o SIFS ( * forno forimm K for Dly K for Dly AGG Data MAC H MAC HS pre PHY o SIFS from (1, (11, and (13, the normalized throughput during a CAP is given by h CAP PL S Data*8 (1 p δ + P + ( p P PL S Data*8 (1 p δ + P + ( p P = PKL S Data*8 (1 p δ + P + ( p P PKL S Data*8 (1 p δ + P + ( p P b S s CAP b S f CAP b S s CAP b S f CAP b S s CAP b S f CAP b S s CAP b S f CAP for No for No for Dly for Dly AGG from (1, we can also calculate the average access delay during a /CAP. (13 ( Performance Evaluation For the performance evaluation, we adopt the following parameters from [11] as shown in table 2. Parameters SIFS MIFS Preamble and PLCP Header Values 2.5 usec 1 usec 9 usec CW min 8 Payload Size Policy Data Rate Control Signal Rate 1~5 KB 3 basic +Dly--AGG policies 1~2 Gbps 48 Mbps Analysis able 2. Parameters values Figures 1 and 2 show the normalized throughput for different payload sizes with different polices with and without aggregation, respectively. We assume an ideal channel conditions for these results. Here, we can observe that No- gives the superior results as most of the time is utilized for data transfer. However, No- policy is not suitable for every application and channel condition. he Dly--

8 AGG policy can achieve somewhat close results to No- policy, as it reduces the unnecessary inter-frame time as well as the header size. Figures 1 and 2 give us the normalized value of throughput at different payload sizes but can t answer for optimum payload size. So, we produce the same results with Gaussian wireless channel model with a given BER rate. Different schemes without aggregation Method Different schemes with aggregation method hroughput (Mbps hroughput (Mbps hroughput (No--1 Gbps hroughput (Imm--1 Gbps hroughput (2-Dly--1 Gbps hroughput (4-Dly--2 Gbps hroughput (No--2 Gbps hroughput (Imm--2 Gbps hroughput (2-Dly--2 Gbps hroughput (4-Dly--2 Gbps Data Packet Payload (KB Fig.1. hroughput versus payload size with different policies Different Schemes without aggregation Method (BER=.2 1 hroughput (No--1Gbps hroughput (Imm--1Gbps 9 hroughput (2-Dly--1Gbps hroughput (4-Dly--1Gbps hroughput (No--2Gbps 8 hroughput (Imm--2Gbps hroughput (2-Dly--2Gbps 7 hroughput (4-Dly--2Gbps Data Packet Payload (KB hroughput (Mbps hroughput (Mbps hroughput ( No- 1 Gbps hroughput ( 2-Dly- 1 Gbps hroughput ( 4-Dly- 1 Gbps hroughput ( No- 2 Gbps hroughput ( 2-Dly- 2 Gbps hroughput ( 4-Dly- 2 Gbps Data Packet Payload (KB Fig.2. hroughput versus payload size with different policies Different Schemes with aggregation Method (BER=.2 18 hroughput (No--1Gbps hroughput (2-Dly--1Gbps 16 hroughput (4-Dly--1Gbps hroughput (No--2Gbps 1 hroughput (2-Dly--2Gbps hroughput (4-Dly--2Gbps Data Packet Payload (KB Fig.3. hroughput versus payload size with different policies Fig.4. hroughput versus payload size with different policies Figures 3 and 4 show the throughput for different payload sizes under a given BER value. It can be seen that an optimal payload size exists for a given BER value. As shown in the mentioned figures the throughput first increases, and then decreases with increasing payload size (even with the aggregation in error prone channels. his is because without the protection of FCS in individual payload frame, a single bit error may corrupt the whole frame which will waste lots of medium time usage and counteract the efficiency produced by an increased payload size. Figure 5 shows the normalized throughput for different BER values with different policies when payload size is set to 1KB. As the BER value increases the throughput decreases. he No- policy with aggregation has larger throughput over large range of BER values than any other policies. o find the effect of n-dly- on bandwidth utilization as well as to find the optimal value of burst size for n-dly- policy, we

9 define the MEB metric in (6. Figures 6 and 7 show the MEB with different burst values under a given BER value. From these figure we can observe that burst size = 4 gives good results in fairly all given payload values. Figure 8 shows the access delay performance for different burst sizes with the aggregation method. Here, we only analyzed n-dly--agg policy to get the upper bound on the delay performance. hroughput Gbps and Payload=1KB No- ( Without Aggregation Imm- 2-Dly- ( Without Aggregation 4-Dly- ( Without Aggregation No- (With Aggregation 2-Dly- ( With Aggregation 4-Dly- ( With Aggregation MEB Without aggregation method (BER=.2 Payload = 1KB- 1Gbps Payload = 2KB- 1Gbps Payload = 3KB- 1Gbps Payload = 4KB- 1Gbps Payload = 5KB- 1Gbps MEB BER Fig.5. hroughput versus BER value with different policies With aggregation method (BER=.2 Payload = 1KB- 1Gbps Payload = 2KB- 1Gbps Payload = 3KB- 1Gbps Payload = 4KB- 1Gbps Payload = 5KB- 1Gbps Burst Size n Delay (usec Burst Size n Fig.6. MEB versus burst size (Data Rate=1Gbps, BER=, n-dly- Payload = 1KB Payload = 2KB Payload = 3KB Payload = 4KB Payload = 5KB Burst Size n Fig.7. MEB versus burst size CAP Analysis Fig.8. Access delay versus burst size Figures 9 and 1 show the normalized throughput over different payload sizes with different policies. Here, we assume no competition between active nodes, as our main focus is to get maximum normalized throughput for each payload size. Figures 11 and 12 show the normalized throughput for different payload sizes with different policies under a given BER value. For each policies, with the increase of the payload size, the throughput first increases, then decreases after the maximal point. his can be explained as follows. In CAP, the time to transmit the payload is only a small portion of the total time used. herefore, when the payload size increases, the transmission efficiency can be increase, but the error probability also increases. he increase of the curves in figures 11 and 12 is because the effect of increased transmission efficiency is more significant than the effect of increased frame error

10 probability, and the decrease of the curves is due to dominant effect of increased frame error probability when payload size further increases. From the above mentioned figures we can find out the optimum payload size value for a given BER value. Here, it is worth to note that normalized throughput performance depends on the number of active stations and backoff window size during a CAP time. Figure 13 shows the normalized throughput for different BER values when payload size is set to 1KB. Different schemes without aggregation method (During CAP time 1 hroughput (No--1 Gbps 11 hroughput ( Imm--1 Gbps hroughput (2-Dly--1 Gbps 1 hroughput (4-Dly--1 Gbps hroughput (No-- 2 Gbps 9 hroughput (Imm--2 Gbps hroughput (2-Dly--2 Gbps 8 hroughput (4-Dly--2 Gbps hroughput (Mbps Data Packet Payload (KB Fig.9. hroughput versus payload size hroughput (Mbps Different schemes with aggregation method ( During CAP ime hroughput (No- 1 Gbps hroughput (2-Dly- Gbps hroughput (4-Dly- 1 Gbps hroughput (No- 2 Gbps hroughput (2-Dly- 2 Gbps hroughput (4-Dly- 2 Gbps Data Packet Payload (KB Fig.1. hroughput versus payload size Different Schemes without aggregation method (During CAP time, BER=.2 7 hroughput (No--1Gbps hroughput (Imm--1Gbps hroughput (2-Dly--1Gbps 6 hroughput (4-Dly--1Gbps hroughput (No--2Gbps hroughput (Imm--2Gbps 5 hroughput (2-Dly--2Gbps hroughput (4-Dly--2Gbps hroughput (Mbps 3 hroughput (Mbps Different Schemes with aggregation method (During CAP time, BER= hroughput (No--1Gbps hroughput (2-Dly--1Gbps hroughput (4-Dly--1Gbps hroughput (No--2Gbps hroughput (2-Dly--2Gbps hroughput (4-Dly--2Gbps 1 Data Packet Payload (KB Fig.11. hroughput versus payload size Data Packet Payload (KB Fig.12. hroughput versus payload size

11 hroughput Gbps and Payload = 1 KB ( During CAP ime No- ( Without Aggregation Imm- 2-Dly- ( Without Aggregation 4-Dly- ( Without Aggregation No- (With Aggregation 2-Dly- ( With Aggregation 4-Dly- ( With Aggregation MEB Without aggregation method (During CAP BER=.2.2 Payload = 1KB- 1Gbps.18 Payload = 2KB- 1Gbps Payload = 3KB- 1Gbps.16 Payload = 4KB- 1Gbps Payload = 5KB- 1Gbps BER Burst Size n Fig.13.hroughput versus BER value with different policies Fig.14. MEB versus burst size.25.2 With aggregation method (During CAP, BER=.2 Payload = 1KB- 1Gbps Payload = 2KB- 1Gbps Payload = 3KB- 1Gbps Payload = 4KB- 1Gbps Payload = 5KB- 1Gbps (During CAP, Data Rate = 1Gbps, BER=, n-dly- Payload = 1KB Payload = 2KB Payload = 3KB Payload = 4KB Payload = 5KB 3.5 MEB.15.1 Delay (msec Burst Size n Fig.15. MEB versus burst size Burst Size n Fig.16. Access delay versus burst siz Similar to duration analysis, Figure 14 and 15 shows the MEB for different burst values under a given BER value. Figure 16 shows the access delay performance for different burst sizes under a given BER value. he access delay performance with different policies is very sensitive to backoff window size and number of active nodes. In this paper, we focused only on n-dly--agg policy to get the upper bound on the delay performance. 4 Conclusions and Future Work In this paper, we presented a performance analysis of WPAN (IEEE MAC for wireless sensor networks. We have extensively studied the different policies in and CAP under a given BER value. he optimal payload size as well as optimal burst size can be determined analytically from the presented analysis. For future work, we want to study the impact of backoff window size on channel utilization and access delay. We hope that the results of this paper will help sensor

12 network designers to easily and correctly provision systems based on IEEE MAC technology. References 1. P /D17.:(C/LM Standard for elecommunications and Information Exchange Between Systems - LAN/MAN Specific Requirements - Part 15.3: Wireless Medium Access Control (MAC and Physical Layer (PHY Specifications for High Rate Wireless Personal Area Networks. Feb Part 15.3: Wireless Medium Access Control (MAC and Physical Layer (PHY Specifications for High Rate Wireless Personal Area Networks. Amendment 1:MAC Sub layer, IEEE Std b, Feb Y. Xiao.: IEEE 82.11n: Enhancements for higher throughput in wireless LANs. IEEE Wireless Communications, December 5. pp K.-W. Chin, D. Lowe.: Simulation study of the IEEE MAC. in: Proc., Australian elecommunications and Network Applications Conference (ANAC, Sydney, Australia, K.W. Chin and D. Lowe.:A Novel IEEE Sharing Protocol for Supporting VBR Streams. in proc. of IEEE ICCCN, H. Chen, Z. Guo, R. Yao, and Y. Li.: Improved Performance with Adaptive Dly- for IEEE WPAN over UWB PHY, IEICE RANS. Fundamentals, VOl-E88-A, No.9, Sep Y. H. seng, E. H. Wu, and G. H. Chen.: Maximum raffic Scheduling and Capacity Analysis for IEEE High Data Rate MAC Protocol. in proc. of IEEE VC, Y. Xiao, X. Shen, and H. Jiang.: Optimal Mechanisms of the IEEE MAC for Ultra-Wideband Systems. IEEE JSAC, Vol.24, No.4, April, G. Bianchi.: Performance analysis of the IEEE distributed coordination function. IEEE JSAC, Vol.18, no-3, pp , Mar.. 1. H.Chen, Z. Guo, R. Y. Yao, X. Shen, and Y.Li.: Perfoamnce Analyis of delyed Acknowledgemnet Scheme in UWB-Based High-Rate WPAN. IEEE transaction on vehicular technology, Vol.55, No.2, March IEEE P82.15 Working Group for Wireless Personal Area Networks (WPANs.: Unified and flexible millimeter wave WPAN systems supported by common mode. DOC: IEEE c, July 9.