Performance of Random Contention PRMA: A Protocol for Fixed Wireless Access

Transcription

1 Int. J. Communiations, Network and System Sienes, 2011, 4, doi: /ijns Published Online July 2011 ( Performane of Random Contention PRMA: A Protool for Fixed Wireless Aess Salman Ali AlQahtani 1, Ahed Alshanyour 2, Ashraf Mahmoud 2 1 Department of Information and Communiations Tehnology, King Fahad Seurity College, Riyadh, Saudi Arabia 2 Computer Engineering Department, King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia salman@kfs.edu.sa, {ahed, ashrafs}@kfupm.edu.sa Reeived May 21, 2011; revised June 24, 2011; aepted July 2, 2011 Abstrat This paper introdues a paket reservation multiple aess (PRMA) with random ontention for fixed wireless aess ommuniations. The performane of PRMA sheme with random ontention is ompared with the performanes of traditional PRMA under permission ontention sheme. The proposed sheme is a simpler ontention mehanism that does not depend on a pre-determined permission probability as PRMA under permission ontention sheme. In this new method, terminals selet the ontention slot uniformly from the pool of remaining free slots in the urrent frame. We evaluate the performane of the new ontention mehanism in terms of various metris inluding maximum number of arried voie alls and paket delays for a given aeptable drop rate of voie pakets. We show that the new mehanism is superior to that of PRMA under permission ontention sheme for loaded systems and is expeted to be insensitive for traffi soure burstiness. Keywords: Contention Sheme, Fixed Wireless, Performane Evaluation, PRMA 1. Introdution Paket reservation multiple aess (PRMA) is designed for a two-way wireless ommuniations network with a star topology. It enables dispersed terminals to transmit paketized information over a shred hannel to a base station. While PRMA ontrols the upstream traffi, the base station broadasts ontinuous binary feedbak messages to the ontended stations on eah slot to indiate availability of these slots for ontention for the subsequent frames. Due to this low ontrol and feedbak messages, PRMA is designed for indoor or short range miroell where the transmission delay are negligible and the feedbak assumed to be available instantaneously. In this paper we fous on voie transmission while PRMA ould be used for integrated voie and data transmission. PRMA uses TDMA tehnology where transmission frame is divided into a set of slots. Terminals have pakets for transmission ontend on these slots. Suessful ontention leads to slot reservation but suessful ontention ours only if one terminal ontends on a given slot. Suessful reservation for a slot is aknowledged by the base station. Collision ours if more than one terminal ontends for the same slot. Terminal that suess in reserving a slot keeps on this reservation for transmitting subsequent pakets, slot reservation is terminated as terminal buffer has no pakets to transmit. To inrease system effiieny PRMA uses speeh ativity detetion. Speeh ativity detetion is used at voie terminals to determine when the speaker is silent or talking. Contention sheme in PRMA is based on permission probability, terminal that has pakets to transmit an ontend on the slot only if it has a permission to ontend, this type of ontention is used to avoid ollision ourrene. For maximum system apaity, permission probability should be small beause inreasing this permission probability leads to exessive number of ollisions. Also reduing the permission probability beyond a ertain value leads to a poor utilization for free slots. For maximum system utilization, Nada and Goodman [1] use equilibrium point analysis to evaluate system behavior. For a partiular example their numerial alulations show that the maximum apaity ould be ahieved with permission probability rages between 0.25 and 0.35 is 37 simultaneous onversations, these simultaneous onversations share a PRMA hannel with 20 slots per frame [2].

2 418 S. A. ALQAHTANI ET AL. All terminals mentor the binary feedbak messages that are broadasted by the base station. Based on these feedbak messages terminals an know in advane the available slots within the transmission frame. Using permission sheme as proposed in [2], all ative stations that have permission to transmit are ontending on the first inoming available slot. This simultaneous ontention will inrease the probability of ollision ourrene. In this work, we propose a new ontention sheme that redues the number of simultaneous ontention on a speifi available slot. The new proposed sheme, random ontention sheme, is based on the fat that available slots are known in advane. Therefore, rather than allowing all ready ative stations to ontend on the inoming available slot, we limit this number of simultaneous ontention by allowing eah ative terminal to selet its expeted available slot randomly from the pool of remaining free slots in the urrent frame. The new sheme still has the maximum apaity that is reahed by the permission ontention sheme. This work is organized as follows. Setion two is a survey of related work to PRMA. In Setion three we desribed the PRMA tehnique as well as the new proposed ontention sheme. In Setion four, we present the simulation model used for performane analysis, and we disuss workload model and parameters used in the simulation. Also, the results of the performane evaluation are presented and disussed. Finally, onlusions are given in Setion five. 2. Related Works Contention tehniques suh as ALOHA and arrier sense multiple aess (CSMA) are not suitable for serving large number of terminals. These tehniques make ineffiient use of the shared transmission medium. Moreover, with inreasing number of terminals throughput goes down and transmission delay inreases. [2] explored PRMA as a tehnique for transmitting over short range radio hannels. The system is proposed as a solution for ineffiient ontention tehniques. There have been many proposals for enhaning the performane of PRMA whih onentrated on improving hannel effiieny more and providing some kind of fairness for data appliations, negleting the variable hannel aess delay problem [3-10]. Goodman et al. previous work extended in [3] by examining the influene of a large number of system variables on PRMA performane. Through omputer simulation they found that with 32 kbps speeh oding and 720 kbps transmission medium PRMA an support up to 37 simultaneous onversations under dropping probability less than In [1], their previous work are extended by using equilibrium point analysis to evaluate PRMA system behavior mathematially. They derive the probability of paket dropping given the number of simultaneous onversations; they also establish onditions for system stability and effiieny. Their numerial alulation show lose agreement with omputer simulation results. In [4], it has been shown that PRMA graefully aept low-rate data terminals with moderate data paket delays and extended the equilibrium point analysis by inluding voie and data traffi in the numerial model. To prevent slots from being wasted in the event of a ollision during aess ontention, [5] proposed a minislotted PRMA protool. The proposed protool is shown to yield improvements over PRMA for both types of traffi, voie and data, at the expense of introduing small amounts of additional lipping to some talk spurts. In an ideal ondition with only speeh traffi, the proposed system supports 41 speeh terminals with less than 1% pakets dropped. To overome the stability problem under heavy load onditions, a modified version, whih is referred to as non-ollision PRMA, NC-PRMA, is proposed in [6]. The protool is based on assigning a pair of dediated ontrol minislots in the all set-up phase for eah terminal, the ontention mehanism required for reservation in the PRMA protool is replaed by an individual reservation proedure. As expeted, paket ollision never ours. The study results also show that the overall performane of the NC-PRMA system is superior to that of the PRMA system. Authors in [7] proposed a hybrid aess protool known, as ontention time-division multiple aess (C- TDMA). C-TDMA shows some features of ontentionbased (slotted-aloha) and reservation-based (PRMA) protools. It has been reommended for use in the uplink of future European multimedia distribution systems. A simple Markov model is proposed to desribe C-TDMA behavior. Their results in terms of throughput and delay under variable traffi onditions indiate that C-TDMA is able to grant optimum throughput/delay figures for typial multiuser systems. Moreover, for a digital speeh senario, a performane omparison with PRMA demonstrates that C-TDMA yields equivalent performane to PRMA in terms of number of users supported by the system with a limited paket dropping rate. In this work, whih is an extension of [9], we propose the use of a simple ontention resolution method for voie terminals that does not depend on a pre-determined permission probability figure as in the original PRMA. In this method, the user selets a slot uniformly from the pool of remaining free slots in the frame. In the results setion we show that the proposed method while produes the same apaity figures as the original method, provides lower paket delays for loaded systems. Finally,

3 S. A. ALQAHTANI ET AL. 419 we expet the new method to be insensitive with respet to traffi burstiness degree. 3. PRMA System Desriptions PRMA is a tehnique for transmitting, over short range radio hannels, a mixture of voie and data pakets. The PRMA protool is organized around time frames with duration mathed to the periodi rate of voie pakets. In eah frame, ative terminals try to reserve free time slot dynamially. PRMA is losely related to reservation ALOHA, R-ALOHA. It is distinguished from R-ALOHA by its response to network ongestion and by its short round-trip transmission time. In R-ALOHA, ongestion auses exessive paket delays. In PRMA, information pakets are disarded if they remain in terminal transmission buffer beyond a ertain time limit. For indoor appliation the round-trip propagation time between terminals and base stations is less than one miroseond. Paket durations typially are µs. The short propagation times allow terminals to learn quikly the results of transmission attempts. In many ases, an aknowledgment message for the urrent time slot an arrive at the terminals before the beginning of the next time slot [1,2]. PRMA is a slotted protool where frame duration, T, is a design variable parameter divided into a set of slots, N, number of these slots depends on both frame size and transmitted paket size. Frame size is related to frame duration and hannel transmission rate, R. Paket size is related to soure bit rate, R s. Terminal generates only one speeh paket during frame time. The amount of soure information per paket is R s T bits and the total length is R s T + H bits. H represents the size of the header. Header ontains routing, synhronization and ontrol information. RT N (1) RT H The speeh ativity detetor is modeled as a two-state Markov proess as shown in Figure 1. The probability of transmission from the talk, TLK, state toward the silene state, SIL, is the probability that the talkspurt with mean duration t1 ends during a time slot of duration. 1exp rt 1 (2) While the probability of transmission from the SIL state to the TLK state is the probability that silenespurt with mean duration t 2 ends during a time slot 1exp t 2 (3) Figure 2 represents a PRMA speeh terminal with N + 2 states. Terminal ould be in one of the N + 2 states, speeh ativity detetor ontrol the transmission of the terminal between SIL state and ontending state, CON. Deteting talkspurt moves the terminal toward CON state and terminal starts ontending on available time slots as pakets be ready for transmission in its transmitting buffer. Deteting silenespurt by speeh ativity detetor moves the terminal into the silene state. The terminal in CON state gets a reservation and start transmit if the following three onditions are hold simultaneously: 1)The ontended slot is unreserved; the probability that the slot is not reserved is (1 r), where r is the probability that the slot is reserved by some other terminal. 2) The terminal has a permission to transmit, transmission permission is p. 3) No other terminals have a permission to transmit u 1 p 1, where represent the number of ontended terminals. Permission for transmission is generated aording to a binary random event generator independently at eah terminal. The permission probability, p, is a system design parameter. A feedbak message from the entral base station is broadasted to all terminals with the end of eah time slot to indiate whether this slot is reserved or still available for reservation in the next frame. By the starting of eah slot, terminal that has pakets and permission start transmitting their first available paket. Base station detets ollision if it gets a disturbed signal due to the transmission of two pakets or more within the Figure 1. Two-state model of a slow speeh ativity detetor. States are talking (TLK) and silene (SIL). Figure 2. Transition State Diagram for Voie Terminals, γ f, represents the probability of transition from state RES0 to state SIL, this probability is equal the probability that the talkspurt ended in the most reent frame.

4 420 S. A. ALQAHTANI ET AL. same time slot. Terminals an detet whether they enounter a ollision or no by deteting the feedbak message broadasted by the base station. If some terminal enounters a ollision it keeps the transmitted paket in its buffer and tries to ontend on other available time slot. Terminal that sueeded in reserving the ontended time slot keeps on this reservation for transmitting all subsequent speeh pakets it generates during the talking spurt. In our new ontention sheme, terminal with ready pakets to transmit selet their slots by running a uniform random generator on available free slots. Collision ould our if more than one terminal ontends randomly for the same time slot. Terminal starts transmitting its speeh pakets as the ontended slot is oming. Collision deteted by the terminal after reeiving the feedbak broadasted message from the base station. Collided terminals start piking new available slots for transmission. This sheme of ontention ould inrease paket queuing delay by deferring paket transmission until the availability of the ontended slot, but it redues the number of ollisions ompared with the permission ontention sheme. Reduing number of ollisions will redue paket queuing delay for heavy traffi systems, beause in these systems the probability of ontending on the same slot by more than one terminal is relatively high. This high probability indues suessive ollisions for inoming slots. Reduing this probability requires reduing permission probability, but reduing permission probability beyond a ertain value results in wasting many available time slots. Speeh pakets are sensitive to the delay; therefore, speeh pakets annot tolerate high queuing delay, a maximum threshold delay, D max is onsidered for pakets. Pakets with queuing delay exeeding this threshold delay dropped from terminal transmitter buffer. Pakets transmission depends on slot reservation; therefore pakets ould enounter exessive queuing delay. Terminals transmitter buffers start dropping the oldest pakets if there is no available spae for reent generated pakets. Buffer size, B, depends on both the maximum threshold delay and the frame duration. B D max T (4) under permission ontention sheme. The number of simultaneous onversations has been used as a running parameter in all experiments; the maximum number of simultaneous onversations that an take plae with p drop 0.01 is adopted as a measure of system apaity, M. p drop is defined as the number of dropped pakets to total number of generated pakets. The Effet of different system parameters on PRMA with random ontention sheme are evaluated and ompared with PRMA under permission ontention sheme. With our new proposed ontention sheme, terminals always have permission for transmission. While in permission ontention sheme terminals are allowed to ontend on the available slots only if they have permission to do that, this permission is generated in eah terminal independently from other terminals. Figure 3 shows that with permission sheme the maximum number of simultaneous onversations is hanged with hanging permission probability. Table 1. PRMA variables. Parameter Notation Units Values Talk Spurt Mean Duration t 1 se 1.0 Silent Spurt Mean Duration t 2 se 1.35 Channel Rate R b/s Soure Rate R s b/s Frame Duration T se Overhead H bit 64 Speeh Ativity Detetor Slow Maximum Delay D max se Dropping Probability p drop 0.01 Permission Probability p 0.3 Conversations M Variable 4. Silulation Results A simulation program is written to evaluate the effiieny of PRMA under the new ontention sheme ompared against the permission ontention sheme. Table 1 provides a summary of system parameters that are used in our simulation program. These parameter values are adopted in [2] study for evaluating PRMA effiieny Figure 3. Maximum number of simultaneous onversation with p drop 0.01 as a funtion of permission probability, p.

5 S. A. ALQAHTANI ET AL. 421 For both low and high permission probability, most of available slots are wasted without reservation. With low permission probability slots are wasted beause terminals have no permission for ontention while with high permission probability, slots are wasted due to ollision between ontended stations, the maximum number of simultaneous onversations reahed is M = 37 when p = 0.3. Using our new approah give us the same maximum number of simultaneous onversations. Our approah is better than the permission approah in terms of the average paket delay under heavy load traffi and number of ollisions in light load traffi. Figure 4 and Figure 5 show these fats. Paket delay is onsidered as the elapsed time between paket generation and the start of paket transmission, taking in aount the limitation on this time it should not exeed the maximum paket delay, D max. Figure 4 shows that for maximum number of simultaneous onversations the average pakets delay is less than that in permission ontention sheme. The high values for the average paket delay under light load traffi are due to the nature of random ontention sheme. In random ontention sheme terminals ould ontend on any available slot from the pool of remaining free slots in the urrent frame, this slot ould be just the next one or ould be the last one in the frame, in average terminals need to wait for 4 mse (frame duration, T = 16 ms). The low average paket delay at high traffi load is good beause the objetive from exploring PRMA system is to ahieve high apaity with minimum paket delay. Figure 5 shows the number of ollisions as a funtion of the number of simultaneous onversations. The number of ollisions is inreased with inreasing the traffi load due to the possibility of ontending on the same slot by many terminals at the same time. Permission ontention sheme solved this problem by reduing the permission probability. Our new proposed sheme redues the number of ollisions by distributing slots ontention Figure 4. Paket average delay p drop 0.01 as a funtion of number of simultaneous onversations. Figure 5. Number of ollisions with p drop 0.01 as a funtion of simultaneous onversations. among all available slots rather than ontending on the same slot. Still our sheme faes ollisions due to the possibility of seleting randomly the same available slot by different terminals. But the ourrene number of these ollisions in low traffi ase is less than those in permission ontention sheme. In ase of high traffi load the number of ollisions starts inreasing rapidly beause all terminals that have data to transmit try to ontend randomly on very little number of slots. A solution for this problem ould be by adopting a dynami ontention sheme that redues the number of ollisions in ase of low number of available slots. The frame duration, T, is determined by the amount of speeh information, R s T bits, in eah paket. There is an optimum value for frame duration beause reduing frame duration means that overhead bits will be the dominant, this overhead will onsume exessive hannel resoures. Also, reduing frame duration redues number of slots N per frame (1). This redution in the number of slots per frame inreases number of ollisions beause all terminals try to ontend on small number of slots. Inreasing frame duration also leads to some other problems, inreasing frame duration dereases the buffer size whih in turn leads to exessive paket drooping (4). Buffer size, B, atually representing the number of time slots, D, whih the paket ould live in the terminal transmitting buffer. Eah speeh terminal ontains a firstin first-out buffer to store pakets that waiting suessful transmission. Eah paket in the buffer has a ounter that reords its age in terms of number of slots; if the paket s age exeeds D slots then this paket must be dropped from the buffer. Therefore, inreasing paket duration will derease paket opportunity for transmission. Dmax N D T (5)

6 422 S. A. ALQAHTANI ET AL. Figure 6 shows the effet of frame duration on maximum number of simultaneous onversations with p drop PRMA with random ontention sheme is very sensitive to frame duration. Inreasing frame duration dereases number of simultaneous onversations. With large frame duration, number of time slots, D, that needed for a paket to stay in buffer is dereased and be less than the average paket delay, 4 mse, this redution in D auses the redution in maximum number of simultaneous onversations. PRMA with permission ontention fae the same senario but its sensitivity to frame duration is less. Normalized apaity, η, is used as a measure for omparing different PRMA onfigurations. Number of simultaneous hannels in TDMA system is adopted as a normalized measure [1]. In perfet TDMA system with no overhead, the number of voie hannels is R /R s whih is equivalent to 22.5 (using the proposed parameter values). Normalized apaity,, is defined as: M 0.01 R R (6) Figure 7 shows the normalized apaity under different number of hannels 1, simulation run for different values of transmission rates, R. Inreasing the transmission rate inreases the normalized apaity. But this inrease in normalized apaity is bounded to 1.64; PRMA annot reah its upper bound 2, 2.09, due to the effets of wasted time slots. Both, random and permission ontention shemes, show the same results. But at low traffi rates random ontention sheme gives better apaity than permission sheme. This improvement in apaity at low traffi load is due to utilizing most of available slots, number of ollisions in random ontention at low traffi rates is smaller than that in permission sheme and therefore, number of wasted slots is redued. In PRMA, pakets that stay in terminals transmitting buffers for a time greater than D max will be dropped beause voie appliations annot tolerate transmission delays. Figure 8 displays the relationship between delay limit, D max, and the number of simultaneous on- versation. For low paket delay limits, most of generated pakets are dropped from the buffer. Both ontention shemes are sensitive to paket delay limit. In random ontention sheme most of the generated pakets are dropped from the terminal buffers with low D max, when D max < s all generated pakets are dropped from the terminal transmitting buffers beause the average paket delay is s as shown in Figure 4, therefore, for good 1 Number of hannels is equivalent to TDMA apaity, R is hanged from 64 kbps to 960 kbps, and therefore, number of hannels is ranged between from 2 to Upper bound ould be reahed in the ideal onditions when there are N R R t t t. no wasted slots. upper bound is. s s Figure 6. Number of simultaneous onversation with p drop 0.01 as a funtion of frame duration. Figure 7. Number of simultaneous onversations per hannel as a funtion of the number of hannels. Figure 8. Conversation per hannel as a funtion of speeh paket delay limit. system apaity, D max must be adjusted arefully to avoid this exessive paket drooping. Permission ontention sheme also annot deal with low paket delay limits, as shown in the Figure 8 normalized system apaity under permission ontention sheme at low paket delay limit

7 S. A. ALQAHTANI ET AL. 423 is less than Conlusions In this work we propose a new ontention sheme, random ontention sheme, for PRMA system and we examined the influene of large number of system variables on PRMA performane. Our results are ompared with PRMA system under permission sheme. Simulation results show that with our new sheme we ould reah the maximum number of simultaneous onversation per hannel as in PRMA under permission ontention sheme. Our approah gives better results in some ases. The main features of our approah are: reduing average paket delay at high traffi rates and reduing number of wasted slots for low traffi rates. Simulation results show that for high traffi ases, number of ollisions inreased rapidly, this rapid inrease in number of ollisions is due to small number of available time slots ompared to the number of ative terminals. These results show that the new ontention sheme ould be improved by introduing some dynami ontention poliy for high traffi rate ases. Dynami ontention poliy will inrease system apaity by reduing the number of ollisions. 6. Referenes [1] S. Nanda, D. J. Goodman and U. Timor, Performane of PRMA: A Paket Voie Protool for Cellular Systems, IEEE Transations on Vehiular Tehnology, Vol. 40, No. 3, August 1991, pp doi: / [2] D. J. Goodman, R. A. Valenzuela, K. T. Gayliard and B. Ramamurthi, Paket Reservation Multiple Aess for Loal Wireless Communiations, IEEE Transations on Communiations, Vol. 37, No. 8, August 1989, pp doi: / [3] D. J. Goodman and S. X. Wei, Effiieny of Paket Reservation Multiple Aess, IEEE Transations on Vehiular Tehnology, Vol. 40, No. 1, February 1991, pp doi: / [4] S. Nanda, Analysis of PRMA: Voie Data Integration for Wireless Networks, Proeedings of IEEE Global Teleommuniations Conferene, San Diego, Deember 1992, pp [5] F. Barelo and A. Ramon, Minislotted Paket Reservation Multiple Aess Protool for Integrated Voie-Data Transmission, Proeedings of Singapore ICCS 94 Conferene, Singapore, Vol. 1, November 1994, pp [6] J. -H. Wen and J.-W. Wang, A New Protool For Wireless Voie Communiations-Non-Collision Paket Reservation Multiple Aess, Proeedings of the 6th IEEE International Symposium on Personal Indoor and Mobile Radio Communiations, Toronto, Vol. 2, September 1995, pp doi: /pimrc [7] G. Pierobon, A. Zanella and A. Salloum, Contention- TDMA Protool: Performane Evaluation, IEEE Transations on Vehiular Tehnology, Vol. 51, No. 4, July 2002, pp doi: /tvt [8] J.-H. Wen, Y.-S. Chen, T.-S. Chang and J.-K. Lain, Theoretial Analysis and Performane of NC-PRMA Protool for Multihannel Wireless Networks, Wireless Personal Communiations, Vol. 39, No. 2, 2006, pp doi: /s [9] A. Al-Shanyour, A. A. Mahmoud, T. Sheltami and S. A. AlQahtani, Paket Reservation Multiple Aess (PRMA) with Random Contention, Proeedings of ACS/IEEE International Conferene on Computer Systems and Appliations, Amman, May 2007, pp [10] N. Wattanamongkhol, W. Srihavengsup, P. Vara-urairat, S. Siwamogsatham and L. Wuttisiittikulkij, Analysis of a Multiple-Token Contention Sheme for Broadband Wireless Aess Networks, Proeedings of IEEE International Conferene on Communiations Workshops, Beijing, May 2008, pp doi: /iccw