IEEE COMMUNICATIONS LETTERS, VOL. 11, NO. 4, APRIL 2007 313
A Quorum-based Mechanism as an Enhancement to
Clock Synchronization Protocols for IEEE 802.11 MANETs
Shu-Min Chen, Sheng-Po Kuo, and Yu-Chee Tseng
Abstract— In wireless mobile ad hoc networks (MANETs), it is
essential that all mobile hosts (MHs) are synchronized to a com-mon clock to support the power-saving (PS) mechanism. Many protocols have been proposed for clock synchronization in IEEE 802.11 MANETs. However, it is practically impossible for any protocol to completely solve the asynchronism problem especially when connectivity is achieved by multi-hop communication or when a network could be temporarily disconnected. In this work, we propose a quorum-based mechanism, which includes a new structure of beacon intervals for MHs to detect potential asyn-chronous neighbors and an enhanced beacon transmission rule to assist clock synchronization protocols to discover asynchronous neighbors within bounded time. The proposed mechanism should be regarded as an enhancement to existing clock synchronization protocols. Our simulation results show that the mechanism can effectively relieve the clock asynchronism problem for IEEE 802.11 MANETs.
Index Terms— Mobile ad hoc network, power saving
mecha-nism, wireless network, clock synchronization.
I. INTRODUCTION
I
N wireless mobile ad hoc networks (MANETs), it is essential that all mobile hosts (MHs) are synchronized to a common clock to support the power-saving (PS) mechanism. In IEEE 802.11 PS mechanism, each MH wakes up at the beginning of a beacon interval to exchange messages. Through message exchanging, MHs can schedule communications for the current beacon interval. If a MH is not scheduled for any communication activity, it can go to the doze mode for the rest of the beacon interval. In order to exchange messages properly, MHs’ beacon intervals should be synchronized. Otherwise, PS mechanism will not function well.To fulfill the requirement of clock synchronization, IEEE 802.11 specifies a distributed Timing Synchronization
Func-tion (TSF) for ad hoc networks. Since an ad hoc network
has no infrastructure to provide a centralized synchronous mechanism, MHs will compete with each other to broadcast their timing information through beacons. Each MH receiving beacons will adjust its clock to synchronize with the sender if its current time is slower than the timestamp in the beacon. Note that the clock is unchanged if its current time is faster. Manuscript received November 30, 2006. The associate editor coordinating the review of this letter and approving it for publication was Dr. Rohit Nabar. S. M. Chen and S. P. Kuo are with the Department of Computer Science, National Chiao-Tung University, Taiwan (e-mail: [email protected]).
Y. C. Tseng is with the Department of Computer Science, National Chiao-Tung University, Taiwan, and the Department of Information and Computer Engineering, Chung-Yuan Christian University, Taiwan. Y. C. Tseng’s research is co-sponsored by Taiwan’s MoE ATU Program, by NSC under grant numbers 93-2752-E-007-001-PAE, 95-2623-7-009-010-ET, 95-2218-E-009-020, 95-2219-E-009-007, 94-2213-E-009-004, and 94-2219-E-007-009, by Realtek Semiconductor Corp., by MOEA under grant number 94-EC-17-A-04-S1-044, by ITRI, Taiwan, and by Intel Inc.
Digital Object Identifier 10.1109/LCOMM.2007.061963.
The TSF mechanism is quite enough for small and static ad hoc networks. However, it is pointed out in [1] that IEEE 802.11 TSF has the scalability problem. As the number of MHs increases, the contention among beacons may get very serious, causing beacon loss and thus the clock asynchronism problem. A simple protocol called ATSP is then proposed, which gives faster MHs higher priorities to transmit their beacons. Most existing clock synchronization protocols are mainly designed for fully-connected MANETs. For multi-hop MANETs, several protocols, such as Automatic
Self-time-correcting Procedure (ASP) [2], have been proposed.
It is to be noted that the above protocols all aim at minimizing the maximum clock drift among MHs. If two MHs’ wake-up schedules do not overlap with each other, these protocols can not re-synchronize them. In fact, it is practically impossible for any synchronization protocol to completely solve the asynchronism problem especially when connectivity is achieved by multi-hop communication or when a network could be temporarily disconnected. We will give several examples to prove this argument. MHs’ mobility may further aggravate the clock asynchronism problem. Unable to find asynchronous neighbors will hurt network connectivity and thus communication performance of a network.
This paper proposes a new quorum-based mechanism which can serve as an enhancement to existing clock synchronization protocols and thus should cooperate with one to relieve the clock asynchronism problem. The concept of quorum is borrowed from [3] to guarantee detection of asynchronous neighbors. However, [3] does not try to synchronize MHs’ clocks. Instead, a MH tries to predict other MHs’ clocks by keeping their clock differences and thus needs to deliver buffered packets at right time, which is inefficient in terms of network throughput.
II. PROBLEMDEFINITION ANDBACKGROUNDS In the PS mechanism of the IEEE 802.11 ad hoc mode, a beacon interval consists of a beacon window, an ATIM window, and a DATA window (Fig. 1(a)). A PS MH only needs to wake up during the beacon window and check delivery requests during the ATIM window. The clock asynchronism
problem occurs when the time difference between any two
neighboring MHs is larger than the length of one beacon window. In Fig. 1(a), A can hear beacons sent by B, but
B may not be able to to hear A’s beacons and thus get
synchronized with A. The clock asynchronism problem may remain unsolved until the amount of time drift betweenA and
B is a multiple of one beacon interval (refer to Fig. 1(b)). To
see how serious this problem is, suppose that A is faster than
B by 20µs per beacon interval in Fig. 1. Assuming that the 1089-7798/07$25.00 c 2007 IEEE
314 IEEE COMMUNICATIONS LETTERS, VOL. 11, NO. 4, APRIL 2007
Beacon Window ATIM Window Beacon Interval k Beacon Interval k+1
Beacon Interval k
Beacon Beacon Beacon Beacon
A
B
Beacon Interval m Beacon Interval m+1
Beacon Interval m A B Beacon (a) (b) Beacon DATA Window 100000-1240-16000+1240 =84000 us
Fig. 1. (a) An example of the clock asynchronism problem when MHB is slower than MHA, and (b) resynchronization after B catches up with A’s beacons.
(a) (b)
Fig. 2. An example where clock asynchronism occurs when two disconnected components of a MANET meet each other.
lengths of a beacon interval, an ATIM window, and a beacon window are 100000µs, 16000µs, and 1240µs, respectively (based on the IEEE 802.11 DSSS recommendation), it will take 100000−1240−16000+124020 = 4200 beacon intervals (= 420 seconds) to move from the scenario in Fig. 1(a) to the scenario in Fig. 1(b).
Most existing clock synchronization protocols aim at mini-mizing the maximum clock drift among MHs in a connected and initially synchronized network. It lacks a mechanism to get MHs synchronized when their clocks have seriously drifted away. In a MANET, mobile MHs may be temporarily partitioned into multiple groups. During this period synchro-nization between groups is impossible. For example, Fig. 2(a) illustrates two groups of mobile MHs, each being perfectly clock synchronized but mutually asynchronous. When these two groups merge into one, as shown in Fig. 2(b), MHs in these two groups may not be able to discover each other, and thus the network remains disconnected, which is clearly harmful. Even if a MANET remains connected, a small clock drift per hop may accumulate into a large amount of drift after multiple hops. As shown in Fig. 3, if neighboring MHs’ clock drift is 15 of one beacon window, MHs A and F , which are separated by5 hops, may still remain out-of-synchronization. Therefore, whenF moves to A’s communication range, they may not discover each other, making the network layer mis-takenly interpret thatA and F are quite far away.
III. THEPROPOSEDQUORUM-BASEDMECHANISM Our goal is to design an enhancement that can co-work with existing clock synchronization protocols to relieve the clock asynchronism problem. The basic idea is to use a
C D A E F A Beacon Interval k Beacon Interval k+1 B C D E F Beacon
Window WindowATIM
B
Fig. 3. An example where clock asynchronism causes the network lose the communication link between MHsA and F .
“quorum” concept to schedule MHs’ wake-up time to ensure that a MH can always detect any nearby asynchronous MH within bounded time. Once such MHs are detected, our beacon transmission rule will help synchronize their clocks. The
grid quorum is first proposed in [4] for a MH to discover
asynchronous neighbors. Basically, it requests a PS MH to wake up in only necessary beacon intervals for the discovery purpose. Compared to a random wake-up approach, it can guarantee to discover neighbors within bounded time.
A. Structure of Beacon Intervals
A grid quorum is a 2DN ×N matrix such that each quorum contains a random column and a random row of the matrix. Clearly, the intersection of any two quorums is non-empty. Given a quorum, a MH will group its beacon intervals such that N2 consecutive beacon intervals constitute one group. In each group, itsN2beacon intervals are arranged by row-major in anN ×N grid. The 2N −1 beacon intervals in the selected column and row are quorum intervals, and the remainingN2− 2N + 1 beacon intervals are non-quorum intervals.
Each quorum and non-quorum interval is divided into three parts: beacon window, ATIM window, and DATA window. Beacon transmission rules during a beacon window will be defined in the next section. A MH’s behaviors during ATIM and DATA windows are the same as those defined in IEEE 802.11 except that it has to stay awake throughout the whole DATA window during a quorum interval. Fig. 4 illustrates an example of a grid quorum and the structures of quorum and non-quorum intervals. With such a structure, it has been proven in [4] that two neighboring MHs always have two chances to hear each other’s beacons in every N2 beacon intervals, no matter how much their clocks drift away. The value of N is tunable. A smaller N would facilitate clock synchronization, while a largerN can save MHs’ energy. This will be further evaluated in Section IV.
B. Beacon Transmission Rules
Our quorum-based mechanism has to incorporate with an existing clock synchronization protocol (such as [1], [2]). Let
fi(n) be the beacon transmission decision of MH i made by the adopted clock synchronization protocol for then-th beacon interval (fi(n) = 1 means “transmit” and fi(n) = 0 means “don’t transmit”). The actual beacon transmission is controlled by the beacon-reception and the beacon-window processes in Fig. 5.
CHEN et al.: A QUORUM-BASED MECHANISM AS AN ENHANCEMENT TO CLOCK SYNCHRONIZATION PROTOCOLS FOR IEEE 802.11 MANETS 315 N2 Beacon Intervals N2 Beacon Intervals N2 Beacon Intervals N2 Beacon Intervals N2 Beacon Intervals 0 1 2 N-1 N N+1 N+2 2N N N Quorum Interval Non-Quorum Interval Non-Quorum Interval Beacon Window ATIM Window DATA Window
Group 1 Group 2 Group 3 Group 4 Group 5
N2 -N N2 -1 2N+2 2N-1 3N-1 time
Fig. 4. The structures of grid quorum, quorum interval, and non-quorum interval. Receive j,s Beacon Y N BCNT 1? Quorum Interval? N Send Beacon BCNT 1? Y BCNT= C1 BCNT= C2 N N Y Ti> Tj+ BW ? (Ti< Tj- BW) and (Ni> 0)? End i,s Beacon Window BCNT --fi(n) = 1? Y Y N Y N End
Fig. 5. (a) Beacon-reception process and (b) beacon-window process for MHi.
The beacon-reception process (Fig. 5(a)) is triggered when MHi receives a beacon from MH j. Let Tibe the current time of i, Tj be the timestamp in j’s beacon, BW be the length of a beacon window, andNi be the number of neighbors of
i. MH i will compute beacon counter BCNT as follows: 1) If Ti > Tj + BW , this means that i and j are
out-of-synchronization. Since j and its neighbors may also remain out-of-synchronization withi, i will set BCNT =
C1 to enforce itself to sendC1 beacons in the next C1
beacon windows.
2) If Ti < Tj − BW , i and j are also out-of-synchronization. If i has any neighbor (i.e., Ni > 0), it may also be out-of-synchronization withj. So i will setBCNT = C2to enforce itself to sendC2 beacons in the nextC2 beacon windows.
BCNT is to enforce more beacon transmissions to synchro-nize potential out-of-synchronization neighbors. We recom-mend C1 and C2 to be set to at least N because an out-of-synchronization MH will enter a quorum interval (and thus stay awake in the whole interval) at least once in the next
N beacon intervals. The beacon-window process for MH i
(Fig. 5(b)) is triggered when a beacon window arrives. In any of the following events, MHi will try to transmit a beacon.
1) A quorum interval arrives.
2) BCNT > 0 (in which case BCNT will be decreased by 1).
3) fi(n) = 1.
IV. SIMULATIONS
To verify the effectiveness of the proposed mechanism, a 3000 × 3000 square meters field with 500 random MHs is
22 0.25 0.2 0.15 0.1 0.05 ∞ 32 20 10 5 4 3 2 22 0.25 0.2 0.15 0.1 0.05 Number of Links Quorum Size N
Average Number of Out-Of-Synchronization Links
TSF ASP 580 45 40 35 30 25 20 15 10 5 ∞ 32 20 10 5 4 3 2 580 45 40 35 30 25 20 15 10 5
Duration (Beacon Intervals)
Quorum Size N
Average Duration of Out-Of-Synchronization Links TSF
ASP
Fig. 6. Comparison of the average number of out-of-synchronization links in a beacon interval and the average duration when an out-of-synchronization link appears under different quorum sizes.
simulated. Each MH roams around by the random way-point model with a speed of 0 ∼ 5 m/s and a pause time of 20 seconds. The communication range of each MH is identical and is equal to 250 meter. The lengths of a beacon interval, an ATIM window, and a beacon window are 100000µs, 16000
µs, and 1240 µs, respectively. Each MH’s clock speed is
uniformly distributed in [0.9999s, 1.0001s], where s is the standard speed.
We simulate the standard TSF and the ASP [2] with and without our enhancement. Fig. 6 compares the average number of out-of-synchronization links in a beacon interval and the average duration when an out-of-synchronization link appears under different cases. Note that N = ∞ means that no quorum-based mechanism is applied. We can observe that our enhancement can substantially decrease the number of out-of-synchronization links and accelerate the neighbor discovery process.
REFERENCES
[1] L. Huang and T.-H. Lai, “On the scalability of IEEE 802.11 ad hoc networks,” in Proc. ACM Int’l Symp. on Mobile Ad Hoc Networking and
Computing 2002, pp. 173–182.
[2] J.-P. Sheu, C.-M. Chao, and C.-W. Sun, “A clock synchronization algorithm for multi-hop wireless ad hoc networks,” in Proc. 24th Int’l
Conf. on Distributed Computing Systems 2004, pp. 574–581.
[3] J.-R. Jiang, Y.-C. Tseng, C.-S. Hsu, and T.-H. Lai, “Quorum-based asynchronous power-saving protocols for IEEE 802.11 ad hoc networks,”
ACM Mobile Networking and Applications, vol. 10, no. 1-2, pp. 169–181,
2005.
[4] Y.-C. Tseng, C.-S. Hsu, and T.-Y. Hsieh, “Power-saving protocols for IEEE 802.11-based multi-hop ad hoc networks,” in Proc. IEEE
