Shih-Fan Chou1, Jen-Hsi Liu1, Hsi-Lu Chao1, Tzu-Chi Guo1, Chia-Lung Liu2, and Feng-Jie Tsai2
1Department of Computer Science, National Chiao Tung University, Hsinchu, Taiwan
2Information & Communications Research Labs, Industrial Technology Research Institute, Hsinchu, Taiwan
Abstract—The IEEE 802.16 standard is a promising technology for 4G mobile networks. Though supporting versatile service classes, best effort (BE) service class is expected to dominate WiMAX networks, due to operational simplicity. One of bandwidth request mechanisms that subscriber stations (SS) can utilize to issue bandwidth requests (BW-REQ) for BE connections is contention-based random access. An SS starts a timer T16 when transmitting a BW-REQ. If getting a grant before timer expiration, the SS transmits data packets at the allocated time slots; otherwise it performs truncated binary exponential backoff process for BW-REQ retransmission. The default value of T16 is one frame time. However, T16 impacts on contention and request collision significantly. In the paper, we develop an analytical model for T16 timer setting. Besides, we derive analytical expressions for the average number of tries per BW-REQ and the average packet delay. We compare the theoretical results of fixed and adjustable timers. The results show that adjusting timer reduces both the number of collision and the average packet delay.
Keywords－WiMAX, best effort, bandwidth request, contention I. INTRODUCTION
IEEE 802.16 protocol has been standardized for metropolitan broadband wireless access (BWA) systems, and it is a viable technology to be used for connecting local area networks (e.g., IEEE 802.11-bassed WLAN) to the Internet, due to the characteristics of high transmission rate and flexible quality-of-service (QoS). [1]. The IEEE 802.16 MAC layer supports a mandatory PMP architecture, which consists of a base station (BS) serving a number of subscriber stations (SS).
There are two types of duplex scheme, i.e. FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing). In this paper, we focus on TDD mode. TDD mode requires only one channel for transmitting downlink (DL) and uplink (UL) sub-frames at two distinct time slots. Moreover, the DL and UL ratio can be adjusted dynamically.
In order to support multimedia services, the IEEE 802.16 standard [1][2] defines five service classes to accommodate versatile QoS-demand applications (such as VoIP, and MPEG video). These service classes are unsolicited grant service (UGS), extended real-time polling service (ertPS), real-time polling service (rtPS), non-real-time polling service (nrtPS), and best-effort (BE) service. Due to the fact that ―how to perform resource reservation to meet applications’ QoS demands‖ is not within the scope of the standard, it is possible that even VoIP flows would be treated as BE service class.
Therefore, in this paper, we focus on the BE service class.
A BS has the full control of slot allocation. To avoid collisions, SSs should get permission before their data transmission. According to the IEEE 802.16 standard, such an
exclusive channel access is achieved by requiring SSs to send bandwidth requests first. For this purpose, the IEEE 802.16 standard specifies three bandwidth request mechanisms:
contention-based random access and contention free-based polling are two suggested approaches, and piggyback mechanism is optional. These three request mechanisms are applicable to BE service class, and our focus is on the contention-based approach.
The random access contention resolution adopted in WiMAX is based on a truncated binary exponential backoff scheme without carrier sensing. Before each attempt of BW-REQ transmission, an SS randomly selects a backoff timer from [0, Wi-1], where Wi is the contention window size of the ith retry. The backoff time indicates the number of slots that the SS should wait before its BW-REQ transmission. For the first attempt, the contention window size is the minimum value Wmin; the window size after the ith retry is 2iWmin. The window size keeps doubling till it reaches the maximum value Wmax=2rWmin, where r is the maximum backoff stage. For a BW-REQ, an SS can try at most 16 times. Both Wmin and Wmax are defined by BSs, while the WiMAX standard does not provide optimal/suggested values.
When using contention, no explicit acknowledgment (ACK) frame is sent back to indicate whether a bandwidth request (BW-REQ) message is successfully transmitted or not. Instead, a timeout T16 is set to determine whether requiring retransmission or not. The default setting of T16 is one frame time. An illustrative example of contention-based bandwidth request mechanism is shown in Fig. 1. BW-REQs are sent in the contention period of a frame (Fig. 1-○1), and T16 is set simultaneously (Fig. 1- ○2 ). If a grant is given within T16 timeout (Fig. 1-○3), the SS stops contention resolution and use the allocated bandwidth for uplink transmission (Fig. 1-○4 ).
Otherwise the SS believes that its BW-REQ was corrupted, and then restarts a contention resolution process. The SS randomly selects a backoff timer (Fig. 1-○5), and counts down that timer.
When the timer is zero, the SS retransmits the BW-REQ (Fig.
1-○6), and same processes repeat.
Figure 1 Illustration of contention-based bandwidth request mechanism
Recent research of request mechanisms include [3][4][5][6][7]. In [3], the authors conclude that the best size of contention period is (2N-1), and N is the number of SSs.
However, upon heavy traffic load, the number of data slots of an UL subframe decrease as N increases, and a BS may not issue grants to all received BW-REQs. For those refused and collided BW-REQs, the SSs will run the contention resolution mechanism again, and thus delay time increases.
In [4], the authors introduce a new algorithm, called Multi-FS-ALOHA, which divides the contention period into two parts. The first is used by SSs to issue first-try BW-REQs, while the second part is dedicated for retransmission of BW-REQ messages. These two parts are dynamically fixed on a frame by frame basis. The drawback of [4] is that it requires a dedicated feedback channel for operation.
A modified contention resolution process is proposed in [5]
to improve the system performance. Its main idea is assigning different initial window sizes to different scheduling classes.
However, based on the presented simulation results, this algorithm performs similarly to the contention mechanism defined in the standard.
An analytical model of the contention-based bandwidth request mechanism, defined in [1], in a saturated WiMAX network was developed in [6][7]. [8] took the number of contending SSs into account to determine the optimal window size.
Briefly summarizing the introduced literature, performance of the contention-based request mechanism can be improved by (1) reducing the collision probability, (2) dynamically adjusting the contention period according to the number of SSs, (3) assigning different minimum contention window sizes to service classes, and (4) integrating/implementing both piggyback and contention mechanisms. However, these solutions may incur the problem of compatibility.
Two possible reasons that a BW-REQ cannot be granted and need retransmission are: collision, and insufficient UL data slots. The former is due to multiple BW-REQs are transmitted at the same contention slot; the latter is due to the UL data slots cannot accommodate the total demand of received BW-REQs.
However, SSs cannot identify the exact reason why they do not get resource grants, and just perform contention resolution procedure. Upon heavy traffic load, more contentions in a fixed contention period results in more collisions and worse system performance. Thus our idea is to dynamically adjust T16 timeout. BW-REQs may wait longer before perform contention resolution process. The objective of this paper is to develop an analytical mode for T16 derivation.
The rest of this paper is organized as follows. The analytical model of timeout derivation is introduced in Section II. Numerical results are presented and discussed in Section III.
This paper is concluded in Section IV.
II. ANALYTICAL MODEL
In this section, we explain the developed analytical model.
Since we focus on the retransmission caused by insufficient UL bandwidth, T16ib is used to represent the desired timeout.
In addition, we analyze the average tries of a BW-REQ to get a resource grant, and the average packet delay.
In this analytical model, there are N BE connections, and their packet arrival is in Poisson distribution with .
and are the frame time duration and the number of
data slots of a UL subframe. is the percentage of UL data slots which are allocated to BE service class.
A.
In our analysis, we assume there are m slots in a contention period, and each slot can accommodate one BW-REQ message.
Considering a BW-REQ, the probabilities of request collision and insufficient UL bandwidth of its where R is the maximum number of tries.
Given the number of transmitted BW-REQs in frame
We then derive the number of transmitted BW-REQs in a specific frame, say frame . Connections either incurring BW-REQ collision or having packet arrivals in frame (j-1) will where is the probability that an SS has no packet arrivals in
time, and . Through
Figure 2 An illustrative example of packet delay calculation
On the other hand, the number of requests a BS can serve in a UL subframe is
Thus considering the worst case, a successfully transmitted BW-REQ at frame can be served at most after n frames, and
B. Average Number of Tries
Again in the original contention-based bandwidth request mechanism, a REQ is retransmitted when either the BW-REQ experiences a collision, or the BS has no sufficient UL bandwidth to give it a grant. Let X be the number of tries for a BW-REQ to get granted. Since a request can be sent at most times, the average number of tries is
However, in the modified mechanism, the BW-REQ retransmission is only caused by collisions, thus the average number of tries is as listed in (11).
C. Packet Delay
We define the packet delay being the time duration from the first try of a BW-REQ to the time of successful data packet transmission. In the IEEE 802.16 standard, the frame structure of TDD mode includes a downlink subframe (Fig. 2 ○1) and its ith retry. Note that i=0 means the BW-REQ gets grant at its first try. An example to calculate is shown in Fig. 2. U and V are the time durations from sending the first BW-REQ to the end of the contention period (tu) and from the beginning of the uplink data interval (tv) to the time that the first packet has been transmitted, respectively. Using i=0 as an example, the first possibility is that a first-try BW-REQ is successfully transmitted and gets served immediately, its packet delay is
and w.p. stands for ―with probability‖.
The second possibility is this first-try BW-REQ is successfully transmitted but preserved for later grant. In such a case, the packet delay is the same as (12) while with a BW-REQ to get a grant, and it is uniformly distributed within T16ib
time duration.
The last possibility is that this BW-REQ is collided with other requests, and thus the SS doubles the contention window size, randomly selects a backoff value, and retransmits this counter from [0,Wi-1]. According to [10], the probability mass function of random variable is
(14) If this 2nd-try BW-REQ is successfully transmitted while been preserved, it packet delay is
. In general, for the ith attempt, the average packet delay is
Figure 3. settings vs. the number of requests N upon various packet arrival rates
Figure 4 Probabilities of collision and insufficient bandwidth upon various Wmin settings
Table 1. Parameter settings
Parameter Value
8/16/32/64
Maximum backoff stage, r 10
Maximum number of tries, R 16
Number of request slots, m 10
Number of data slots per uplink
subframe, d 20
In this section, we develop a simulation program to validate the analytical model, and compare and discuss the performance of the original and modified contention request mechanisms.
Parameter settings are listed in Table 1.
Fig. 3 shows the settings upon various numbers of successfully transmitted BW-REQs will be preserved longer before getting grants.
In the following experiment, we set be 3, and
setting is based on the results in Fig. 3. We investigated the performance of , as shown in Fig. 4. It is intuitive that both increase as the number of requests increases. Moreover, we observed that when properly setting Wmin (e.g., Wmin=64), is significantly reduced to 1.2×10-3, and maintains at the smallest value among all.
The performance of average number of tries is in Fig. 5 (a) and (b). If a BW-REQ is transmitted successfully to the BS, it may be preserved for future grant. In such a case, the SS does
not need to retransmit this request and thus the number of tries per request reduces, compared with the original contention request mechanism. Note that the average number of tries for both original and modified mechanisms of is more than that of . The reason is that a small contention window size results in a high collision probability.
(a) Original
(b) Modified
Figure 5 The performance of average number of tries of the two contention-based bandwidth request mechanisms
(a) Original
(b) Modified
Figure 6 The performance of mean packet delay of the two contention-based bandwidth request mechanisms
Fig. 6 depicts the mean packet delay of both request mechanisms as the number of requests increases from 10 to 50, upon various Wmin settings. For both mechanisms, when given a Wmin, a large N value results in long delay due to high collision probability and more retries. On the other hand, for a specific N value, the window size of each backoff stage increases, and the average packet delay increases accordingly.
The reason is that when collision occurs, the range of the backoff value becomes larger (0 to Wi -1). An SS is delayed much more frames when using a larger backoff value. The mean packet delay of the modified mechanism is significantly smaller than that of the original mechanism. The reason is that the modified request mechanism preserves successfully transmitted BW-REQs at most (n+1) frames without performing binary exponential backoff process and thus the contention window size is intact. Therefore, it has rather small delay, compared to the original mechanism.
IV. CONCLUSION
In this paper, focused on BE service class and contention-based request mechanism, we developed an analytical model to derive a theoretical T16ib timeout. Dissimilar to the original
contention-based request mechanism that all unsuccessfully transmitted BW-REQs must perform the truncated binary exponential backoff process, the modified mechanism achieves reduction of collisions and tries by adjusts timeout properly for those successfully transmitted BW-REQs while cannot get grants in the next frame. The modeled timeout is a function of (1) number of BE connections, (2) traffic load, (3) retransmission, (4) collision probability, and (5) bandwidth insufficient probability. Numerical results showed that a suitable timeout does reduce the number of tries, and the average packet delay. Since the failure probability of transmitting REQ decreases and the probability of a BW-REQ being hold increases, the number of tries is reduced. In addition, the range of the backoff value grows exponentially when retry occurs. An SS does not need to wait for the backoff counter counting down to zero for BW-REQ transmission when the BW-REQ is hold by the BS. The average packet delay is lower accordingly. From the numeral and simulation results, when the size of initial contention window approaches the number of contention slots, we could get better average packet delay performance. In our case, we suggest that the contention window size is 8.
ACKNOWLEDGMENT
This work was supported in part by NCTU-MTK Research Center under grant 99Q583, in part by National Science Council under grant NSC 99-2219-E-009-013- and in part by Ministry of Economic Affairs and Industrial Technology Research Institute under grant 99-EC-17-A-03-01-0620.
REFERENCES
[1] IEEE Std. 802.16-2004, ―Local and Metropolitan Area Networks, Part 16: Air Interface for Fixed Broadband Wireless Access Systems‖.
[2] IEEE 802.16e-2005, IEEE Standard for Local and metropolitan area networks Part 16: Air Interface for Fixed and Mobile Broadband Wireless Access Systems Amendment 2: Physical and Medium Access Control Layers for Combined Fixed and Mobile Operation in Licensed Bands and Corrigendum 1, 2006.
[3] Taleb T., Fernandez J.C., Hashimoto K., Nemoto Y., Kato N., ―A Bandwidth Aggregation-aware QoS Negotiation Mechanism for Next-Generation Wireless Networks‖, IEEE Global Telecommunications Conference, November 2007, pp.1912-1916.
[4] Lidong Lin, Bo Han and Lizhuo Zhang, ―Performance Improvement using Dynamic Contention Window Adjustment for Initial Ranging in IEEE 802.16 P2MP Networks‖, IEEE Wireless Communications &
Networking Conference (WCNC), 2007, pp.11-15.
[5] Jianhua He, Ken Guild, Kun Yang, and Hsiao-Hwa Chen, ―Modeling Contention Based Bandwidth Request Scheme for IEEE 802.16 Networks‖, IEEE Communications Letters, Volume 11, August 2007 pp.689-700.
[6] Vinel A., Ying Zhang, Qiang Ni, Lyakhov A., ―Efficient Request Mechanism Usage in IEEE 802.16‖, Global Telecommunications Conference, December 2006, pp.1-5.
[7] Wenyan Lu, Weijia Jia, Wenfeng Du, Lizhuo Zhang, ―Performance analysis of the contention resolution scheme in IEEE 802.16‖. Journal of Software, Volume 18, No. 9, pp.2259-2270, 2007.
[8] Sung-Min Oh, Jae-Hyun Kim, ―The Analysis of the Optimal Contention Period for Broadband Wireless Access Network‖, Pervasive Computing and Communications Workshops, March 2005, pp.215-219..
[9] Giuseppe Bianchi, Luigi Fratta, and Matteo Oliveri, ―Performance evaluation and enhancement of the CSMA/CA MAC protocol for 802.11 wireless LANs,‖ in Proc. IEEE PIMRC, Taipei, Taiwan, Sept. 1996, pp.
392–396.
[10] Hai L. Vu, Sammy Chan, and Lachlan L. H. Andrew, ―Performance analysis of Best-Effort Service in Saturated IEEE 802.16 Networks,‖
Proc. IEEE Vehicular Technology, Volume 59, No. 1, January 2010, pp.460-472.
[11] Q. Ni, L. Hu. "An Unsaturated Model for Request Mechanisms in WiMAX". IEEE Communications Letters, Vol. 14, No. 1, Jan 2010, pp.
45-47.
[12] Q. Ni, A. Vinel, Y. Xiao, A. Turlikov, T. Jiang. "Investigation of Bandwidth Request Mechanisms under Point-to-Multipoint Mode of WiMAX Networks". IEEE Communications Magazine, Vol. 45, No. 5, May 2007, pp. 132-138.
1
「ACM Symposium on Information, Computer and Communications Security
（ASIACCS）國際學術會議」
出國報告書
報告人： 交通大學謝續平
日期：2010 年 04 月 30 日
2
一、 出國目的
ACM Symposium on Information, Computer and Communications Security （ASIACCS）為 ACM Special Interest Group on Security, Audit, and Control (SIGSAC) 所贊助與主辦的兩大頂尖會議之一，
接受率約為 10％一頂尖會議為 ACM Conference on Computer and Communications Security (CCS)， 接受率也約為 10％。本人擔任 ACM ASIACCS steering committee chair, 負責推動該會議，並且召 集 steering committee meeting，遴選每年執行單位。此次參加該 國際學術會議，並審查 2011 年主辦單位進度，與 2012 年主辦國家與 單位，並討論會議場地與籌辦流程。
二、 行程
參加 ACM Symposium on Information, Computer and Communications Security 擔任 Steering Committee Chair。
4/9 Taipei – Beijing
4/10 受 ACM ASIACCS steering committee member 以及 Mozilla
Online Ltd. CEO Li Gong 博士邀請訪問 Mozilla Online Ltd. (該公司為
開發 Firefox web browser 的公司，Firefox 瀏覽器為全球最受歡迎的瀏
覽器之一)
3
4/12 受大會以及 Chinese Academy of Sciences, Deputy Director Jiwu Jing 邀請訪問中科院並演講 “Cloud Computing Security”
4/13-16 ACM Symposium on Information, Computer and Communications Security 會議
4/17-18 ASIACCS steering committee 會議擔任主席 4/19 返台
三、 出國人員：
謝續平現任交通大學資訊工程系教授暨 TWISC＠NCTU 主任，曾任 交通大學資訊工程系系主任、交通大學計算機與網路中心主任、中華 民國資訊安全學會理事長，現在擔任 IEEE Tran. On Dependable and Secure Computing 、 IEEE Trans. On Reliability 、 Journal of Computer Security 副編輯、IEEE RS Newsletter 總編輯。由於現在 擔任 ACM Symposium on Information, Computer and Communications Security （ ASIACCS ） 推 動 委 員 會 主 席 （ steering committee chair）。負責遴選籌辦國家單位，並督導籌辦進度。
四、 工作內容摘要
由於擔任 ACM Special Interest Group on Security, Audit, and
Control (SIGSAC) 的 推 動 委 員 會 委 員 （ Steering Committee
member），並且擔任 ACM Symposium on Information, Computer and
4
Communications Security （ASIACCS） 推動委員會主席（steering committee chair） ，被 ACM 賦予：
a) 觀察本年度會議執行成果，
b) 審查下年度執行單位籌備現況，
c) 並甄選兩年後會議執行單位。
本次出國為了推動 SIGSAC 的未來發展，赴大陸北京友誼賓館，參加 本年度會議，觀察 ASIACCS 本年度會議主辦單位美國賓州州立大學、
瑞士 ETH、北京中國科學研究院成果，並審查 2011 會議舉辦單位香 港大學、香港城市大學籌備進度，與 2012 年申請舉辦單位上海交通 大學等單位的提案。
此次大會由北京中國科學研究院 Dengguo Feng 主任擔任大會主席，
David Basin([email protected], ETH Zurich, Switzerland) Peng Liu([email protected], Pennsylvania State University, USA) 擔任議程主席，會議接受率僅約 10％，相較於 IEEE INFOCOMM 等頂 級國際會議的接受率 25%，顯得更為難得。
本次會議前、後分別受到本會議的推動委員會委員 Mozilla 的 CEO Li
Gong 的邀請訪問以及本會議的大會邀請至中國科學研究院演講，而
國際會議後的推動委員會也決議 2012 年的主辦單位延至下次會議
討論。
5
五、 結語
本次大會由有來自全世界三十餘國作者投稿，稿件水準極高，接受率 極低，約為 10％，會議圓滿成功。會議組織與會議議程如下：
CONFERENCE ORGANIZING COMMITTEE
CONFERENCE ORGANIZING COMMITTEE
General Chair
Dengguo Feng
([email protected], Chinese Academy of Sciences, China)
Program Committee Chair
David Basin([email protected], ETH Zurich, Switzerland)
Peng Liu([email protected],
Pennsylvania State University, USA) Local Arrangements
Committee Chair
Jiwu Jing
([email protected], Chinese Academy of Sciences, China)
Publication Chair
Peng Ning
([email protected], NC State University, USA)
Publicity Chair
Jie Li
([email protected], University of Tsukuba, Japan)
Workshop Chair
Dongdai Lin
([email protected], Chinese Academy of Sciences, China)
Tutorial Chair
Zhong Chen
([email protected], Peking University, China)
6
Treasurer
Sencun Zhu
([email protected], Pennsylvania State University, USA)
Web Chair
Ji Xiang
([email protected], Chinese Academy of Sciences, China) Secretary Daren Zha ([email protected])
Zongbin Liu ([email protected])
STEERING COMMITTEE
STEERING COMMITTEE
Shiuhpyng Shieh(Chair), Chiao Tung University, Chinese Taipei
David Basin, ETH Zurich, Switzerland Robert Deng, Singapore Management University, Singapore
Virgil Gligor, Carnegie Mellon University, USA
Hideki Imai, National Institute of Advanced Industrial Science and Technology, Japan Sushil Jajodia, George Mason University, USA
Pierangela Samarati, University of Milan, Italy
Elisa Bertino, Purdue University, USA Mike Reiter, University of North Carolina at Chapel Hill, USA
Li Gong, Mozilla Online Ltd., USA Ninghui Li, Purdue University, USA
Eiji Okamoto, University of Tsukuba, Japan Vijay Varadharajan, Macquarie University, Australia
7
六、會議議程
ASIACCS 2010: Beijing, China Program Sketch
12 April 13:30-18:00 Registration Lobby of Building 2
8:00-8:50 Registration Meeting Room1, Building 8 8:50-9:00 Welcoming Remarks Meeting Room1, Building 8 9:00-10:00 Invited Talk Meeting Room1, Building 8 10:00-10:30 Coffee-break Meeting Room1, Building 8 10:30-12:00 Session 1:Privacy Meeting Room1, Building 8 12:00-13:30 Lunch Cafeteria in Friendship Palace 13:30-15:00 Session 2:Applied Cryptography Meeting Room1, Building 8 15:00-15:30 Coffee Break Meeting Room1, Building 8 15:30-17:00 Session 3: Network Security Meeting Room1, Building 8 17:30-19:00 Dinner Cafeteria in Friendship Palace 19:00-21:00 Steering Committee Meeting
(Steering committee members only )
Second Floor meeting Room,
8:00-8:50 Registration Meeting Room1, Building 8 9:00-10:00 Invited Talk Meeting Room1, Building 8 10:00-10:30 Coffee Break Meeting Room1, Building 8 10:30-12:00 Session 4: Systems Security – I Meeting Room1, Building 8 12:00-13:30 Lunch Cafeteria in Friendship Palace 13:30-15:00 Session 5: Access Control – I Meeting Room1, Building 8
8:00-8:50 Registration Meeting Room1, Building 8 9:00-10:00 Invited Talk Meeting Room1, Building 8 10:00-10:30 Coffee Break Meeting Room1, Building 8 10:30-12:00 Session 4: Systems Security – I Meeting Room1, Building 8 12:00-13:30 Lunch Cafeteria in Friendship Palace 13:30-15:00 Session 5: Access Control – I Meeting Room1, Building 8