Proceedings of the 5th WSEAS Int. Conf. on APPLIED INFOATICS and COUNICATIONS, alta, September -7, 25 (pp24-28 Optimizing Pacet Size via aximizing Throughput Efficiency of AQ on Bluetooth ACL Data Communication Lin LI XIANG, KALU PIYANATH UDAGEPOLA, YANG XIAO ZONG School of Computer Science and Technology Harbin Institute of Technology No.92, West Da-Zhi Street, Harbin, Heilongjiang P CHINA Abstract: Based on the acnowledgement history of the most recently transmitted pacets, an estimate of the channel bit error rate is made, based on which an optimal pacet size can be selected through maximizing the expected throughput efficiency of AQ protocol on Bluetooth ACL data communication lin. Simulation results show that this method wors very well even with a short observation history. Key words: Bluetooth; AQ; ACL; SCO; Throughput Efficiency; BE Introduction Bluetooth (BT [,2,3] is a short-range radio lin intended to be a cable replacement between portable and/or fixed electronic devices. Two types of transmission lins, synchronous connection oriented (SCO and asynchronous connectionless (ACL lins are used. SCO lin is a symmetric point to point lin supporting time-bounded voice traffic. SCO pacets are transmitted over reserved intervals without being polled. ACL lin is a point to multipoint lin between master and all slaves in the piconet and can use all the remaining slots of the channel not used for SCO lin. Bluetooth is a frequency hopping system which can support multiple communication channels in a common area (each channel is defined by a unique frequency hopping sequence. Frequency hopping is used in such a way that the radio is turned to the same frequency for the entire duration of the pacet, but then changes to a different frequency each time it transmits a new pacet or retransmits an erroneous pacet. Since the fading and interference in the new frequency channel will be significantly different than that of the previous one, the use of frequency hopping with AQ provides an effective method of diversity. Automatic epeat equest (AQ protocols are designed to remove transmission errors from data communications systems. When used over relatively high bit-error rate lins (e.g., -5 or higher such as wireless or satellite lins, their performance is sensitive to the pacet size used in the transmission. When too large a pacet size is employed, there is an increased need for retransmissions, while too small a pacet size is inefficient because of the fixed overhead required per pacet. When an AQ scheme is to be used at the lin layer over a relatively high error-rate lin, the pacet size should be chosen based on the error-rate. In this paper, we concern with choosing optimal pacet payload length on the Bluetooth ACL data communication lins, in terms of maximizing the throughput efficiency of AQ protocol based on the acnowledgement history of the most recently transmitted pacets. That is, given the number of pacets that required retransmission, an estimate of the channel bit error rate is made, based on which a pacet size is chosen to maximizes the expected throughput efficiency of the data lin protocol. 2 Bluetooth Data Pacets In Bluetooth, the data on the piconet channel is conveyed in pacets. The general pacet format is shown in Figure. Each pacet consists of 3 entities: the access code, the header, and the payload. In fig., the number of bits per entity is indicated.
Proceedings of the 5th WSEAS Int. Conf. on APPLIED INFOATICS and COUNICATIONS, alta, September -7, 25 (pp24-28 The access code and header are of fixed size: 72 bits and 54 bits respectively. The payload length can range from zero to a maximum of 2745 bits. Different pacet types have been defined. Pacets may consist of the (shortened access code only, of the access code header, or of the access code header payload. Data in Bluetooth can be transmitted asynchronously using ACL pacets. In this paper, we mainly focus on ACL pacets data transfer used in asynchronously connections, such as summarized in Tab., seven ACL pacet types are defined in the Bluetooth. D stands for Data-edium rate, and DH for Data-High rate. D pacets are all 2/3-FEC encoded to tolerate possible transmission errors. Not encoded by FEC, DH pacets are more error-vulnerable, but it can carry more information. Among seven pacets all but one of these pacet types (AUX use a cyclic redundancy chec (CC code and automatic repeat request (AQ for error control. The AUX pacet is similar to DH, but it has no CC code and can t be re retransmitted, so we don t tae the AUX pacet into consideration. Table Summary of Bluetooth ACL data pacets Type Payload User CC FEC Header(bytes Payload(bytes (bits D -7 6 2/3 DH -27 6 no D3 2-2 6 2/3 DH3 2-83 6 no D5 2-224 6 2/3 DH5 2-339 6 no AUX -29 no no Then, the six ACL pacet types have total maximum payload length as follows: D:24 DH:24 D3: DH3:496 D5:2745 DH5:2744. 3 Selecting Optimal Pacet Size 3. Throughput Efficiency of AQ Protocol A protocol performance is usually characterized by many parameters which are defined by the communication system requirements. The most important parameters are the probability of receiving a message without errors and the protocol throughput efficiency. There are several definitions of the protocol throughput efficiency. ost frequently it is defined as the ratio of the mean number of information bits successfully accepted by the receiver to the number of bits that could have been transmitted during the same time interval [4]. To do so we must first derive an expression for the throughput efficiency of the AQ protocol. The expressions derived in this section assume the use of an optimal AQ protocol in that only pacets containing errors are retransmitted. The throughput efficiency of AQ scheme that uses pacets having n bits of information bits is determined by [5] : η = ( / ( n where, the first term /n of the above expression represents the ratio of information bits to total bits in a pacet, and represents the average number of transmission attempts per pacet. Assuming that the AQ scheme retransmits a pacet until the acnowledgement of a successful reception, the average number of attempts,, needed to successfully transmit one pacet is given by [6] : = (-p+2 p (-p+3 p p (-p+ = p where, p is the pacet error rate. So, for a given p, the throughput efficiency of AQ that uses pacets having n bits of information bits is given by: η = ( (- p (3 n 3.2 Choosing Pacet Size based on History of Transmitted Pacets In Bluetooth wireless connection, in order to guarantee reliable transmission an automatic repeat request (AQ mechanism is adopted. That is, the receiving side sends bac special control frame as the acnowledgement or negative acnowledgement (ACK/NACK to the input. In case of losing frame or acnowledgement message, the timer will send out timeout signal when the timer has expired, and to remind other side that some problems have happened and this frame must be retransmitted. At the same time, receiver must be capable of distinguishing (2 2
Proceedings of the 5th WSEAS Int. Conf. on APPLIED INFOATICS and COUNICATIONS, alta, September -7, 25 (pp24-28 between retransmitted and new frame. With an automatic repeat request scheme in Bluetooth specification, D, DH and the data field of DV pacets are transmitted and retransmitted until acnowledgement of a successful reception is returned by the destination (or timeout is exceeded. The acnowledgement information is included in the header of the return pacet, so-called piggy-bacing. To determine whether the payload is correct or not, a cyclic redundancy chec (CC code is added to the pacet. The AQ scheme only wors on the payload in the pacet (only that payload which has a CC. The pacet header and the voice payload are not protected by the AQ scheme. Supposed the BE(bit error rate is b, and, the number of retransmission requests out of the last pacet transmissions, and ' is the payload size used in the previous transmissions. The expected throughput efficiency η ( of AQ protocol pacets is given by [7] : ( p ( p ( p η ( (4 n ( ( p ( p where = b b ( p ' is the pacet error rate of pacets transmission. The pacet error rate, p, for DH pacets is [6,8] : = ( b (5 p ecalling that D pacets are protected by a (, Hamming code (encoded with a 2/3 bloc FEC, i.e., in every bloc, bits are used to encode bits of data, which is capable of correcting one bit error per bit code bloc. The payload is correctly decoded provided that all code blocs contain one or fewer errors. The pacet error rate, p, for D pacets can be approximated as: p 4 / = (( + b( (6 Combining equations (4 (6, we can get η ( for DH and D pacets respectively as: '( ( ( ( ( ( η ( = b [ ] (7 ' '( n ( b ( ( ( 4 / 4 ' / 4 '( / (( + b( ( ( (( + b( (( + b( η ( = b [ ] 4 ' / 4 '( / n ( + + b ( (( b( (( b( (8 where, n=+26. It is now possible to choose the value of, the payload length to be used in future transmissions, so that the ' throughput efficiency of the AQ protocol is maximized. This can be done by choosing the value of that maximizes equation (7 or (8 for a value of that is equal to the number of retransmission requests that occurred during the previous transmissions using the payload size '. 4 Simulation esults Usually, the solution way of the maximization problem for η ( in equation (7 or (8 is difficult; However, for specific values of, and ' equation (7 or (8 can be solved numerically. An optimal value for can now be found using numerical search algorithms. Since the numerical evaluation of this integral is very intensive, a comprehensive search for the optimal value of is not practical. Instead, a restricted search using select values for can be performed. Such a search, for example, can consider values of that are a multiple of ; thereby significantly reducing the complexity of the search. Such a restricted search has little impact on the performance of the protocol since values of that are within bits of the optimal bloc size should result in near-optimal performance. In figure 2, we plot the optimal payload size when a history of 5 previously transmitted bit pacets payload is considered. As can be seen from the fig.2 (a, for DH pacets transmission, when the previous fifty transmissions resulted in no errors the payload length can be maximized to 2744 bits (the maximization throughput efficiency can be gotten at payload length of 32 bits. When one and two errors occurred the payload length can be increased to 2 and 7 bits respectively. When three errors occurred the payload length can be ept at bits and when more than three errors occur the payload length is reduced. As depicted in fig.2 (, for D pacets transmission, when the previous fifty transmissions resulted in no errors the payload length can be maximized to be 2745 bits (the maximization throughput efficiency can be gotten at pacet payload length of 44 bits. When one, two and three errors occurred the payload length can be increased to 25, 9 and 6 bits respectively and when more than three errors occur the payload length is reduced. 3
Proceedings of the 5th WSEAS Int. Conf. on APPLIED INFOATICS and COUNICATIONS, alta, September -7, 25 (pp24-28 33 3 27 24 2 8 2 9 6 3 5 2 25 3 35 4 45 5 45 42 39 36 33 3 27 24 2 8 2 9 6 3 5 2 25 3 35 4 45 5 Throughput Efficiency. 9. 8. 7. 6. 5. 4. 3. 2. 2 3 4 5 6 7 = = 3 = 5 = 8 = = 2 b (bit error rate b (bit error rate (a ( Fig.3 ean throughput efficiency of algorithm for various b. (ad H pacet. (D pacet. Throughput Efficiency.9.8.7.6.5.4.3.2. 2 3 4 5 6 7 = =3 =5 =8 = =2 Let opt ( be the optimal value of chosen for a given value of out of the last pacet transmissions. The efficiency of the AQ protocol with that value of can be computed according to equation (3 combined with equation (5 or (6. Then it can be averaged over the distribution of given b to yield the performance of AQ for a given value of b. Fig.3 shows the mean throughput efficiency of AQ with various values of and b, and a previous pacet payload length of bits. As can be seen from the figure, whether it is D or D pacets transmission, good performance is obtained with a history of just 75 bits payload size (5 pacets at pacet payload size of bits. When b is higher than -5, more of history is required to obtain a reasonable estimate of throughput efficiency for D pacets transmission; but for DH pacets transmission, the situation is quite different, only few or the least history of pacets transfer is required to obtain high throughput efficiency. In the previous fifty pacets transmission with payload length of bits, select optimizing pacet size under different retransmission pacets, fig.4 compares mean throughout efficiency of AQ during DH and D pacets transmission for various b. It is important to note that the performance of AQ is much more vulnerable to DH pacets when b is high. That is, when b is high the use of DH or D pacet type can have a disastrous effect on the throughput efficiency, and D pacets transfer can produce higher throughput efficiency than DH pacets. When b is low, small variations in the throughput efficiency from the different bit error rate b, pacet types (DH/D. So in a high error rate environment, it is better to tae D pacets as data transmission, which accords with the capability of D to tolerate high transmission error rate. Oppositely, in a low error rate environment, it is better to tae DH pacets as data transmission, it is because not decoded by 2/3-FEC DH pacets have relatively higher data transfer rate than D pacets data transmission. 4
Proceedings of the 5th WSEAS Int. Conf. on APPLIED INFOATICS and COUNICATIONS, alta, September -7, 25 (pp24-28.9.8.7.6.5.4.3.2. DH D 2 3 4 5 6 7 5 Conclusion This paper introduces a method to select the optimal pacet payload length used by Bluetooth ACL data lin layer. The throughput efficiency of AQ protocol is given based on retransmission history. So, given pacet transmission record we can choose the pacet size such that the expected throughput efficiency of the AQ protocol is maximized under different channel BEs (bit error rates. Simulation results show that the method wors very well even with a short observation history (5 pacets at size of bits payload, total 75 bits payload transfer. In a high error rate environment, it is better to tae D pacets as data transmission; but in a low error rate environment, it is better to tae DH pacets as data transmission to improve the data transfer rate. IEEE-ussia Conferences icrowave Electronics: easurements, Identification (EIA'2, Novosibirs, ussia, 2, pp. 22-25. [6] L.J. Chen,. Kapoor,.Y. Sanadidi,. Gerla, Enhancing bluetooth TCP throughput via lin layer pacet adaptation, in: Proc. of the 24 IEEE International Conference on Communications (ICC 24, Paris, France, 24. [7] E. odiano, An adaptive algorithm for optimizing the pacet size used in wireless AQ protocols. Wireless Networs, 999, 5: 279-286. [8].C. Valenti,. obert, and J. H. eed, On the throughput of bluetooth data transmissions, in: Proc. of the IEEE Wireless Commun. and Networing Conf., Orlando, USA, 22, 9-23. eferences: [] Bluetooth SIG, Specification of the Bluetooh System, version., Feb.2. [2] J.Harrtsen, The bluetooth radio system, IEEE Personal Communicatios, Feb.2, 7(: 28-36. [3] X. Li, X.Z. Yang, ebluescatter: an Energy-Efficient Ad Hoc Networ Formation Algorithm over Bluetooth, WSEAS Transactions on Information Science and Applications, 25, 8(2, pp.34-45. [4] W. Turin, Throughput analysis of the Go-Bac-N protocol in fading radio channels. IEEE Journal on Selected Areas in Communications, 999, 7(5, pp. 88-887. [5] V.P. Pribylov and G.A. Chernetsy, Throughput efficiency of automatic repeat request algorithm with selective reject in communication lins with great signal propagation delay, in: Proc. of the 3-rd 5