Mobile-Agent-Based Handoff in Wireless Mesh Networks: Architecture ...

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 58, NO. 8, OCTOBER 2009

4565

Mobile-Agent-Based Handoff in Wireless Mesh Networks: Architecture and Call Admission Control Bo Rong, Member, IEEE, Yi Qian, Senior Member, IEEE, Kejie Lu, Senior Member, IEEE, Rose Qingyang Hu, Senior Member, IEEE, and Michel Kadoch, Senior Member, IEEE

Abstract—The wireless mesh network (WMN) has recently emerged as a promising technology for next-generation wireless networking. In the WMN, it is crucial to support mobile users roaming around the network without service interruption. This consideration motivates us to develop an efficient fast handoff approach using distributed computing technology. Particularly, we propose a mobile agent (MA)-based handoff architecture for the WMN, where each mesh client has an MA residing on its registered mesh router to handle the handoff signaling process. To guarantee quality of service (QoS) and achieve differentiated priorities during the handoff, we develop a proportional threshold structured optimal effective bandwidth (PTOEB) policy for call admission control (CAC) on the mesh router, as well as a genetic algorithm (GA)-based approximation approach for the heuristic solution. The simulation study shows that the proposed CAC scheme can obtain a satisfying tradeoff between differentiated priorities and the statistical effective bandwidth in a WMN handoff environment. Index Terms—Call admission control (CAC), genetic algorithm (GA), handoff, mobile agent (MA), wireless mesh network (WMN).

I. I NTRODUCTION

I

N THE last few years, the wireless mesh network (WMN) has drawn significant attention as a fast, easy, and inexpensive solution for broadband wireless access [1]–[3]. However, there are still many technical challenges that we have to overcome before the WMN can fully be deployed. Particularly, it is crucial to provide mobility support in the WMN, because wireless users are free to move to anywhere at anytime [4], [5]. In this paper, we propose a mobile agent (MA)-based handoff approach to address the issue of user mobility. Our approach aims to reduce handoff delay and provide seamless handoff Manuscript received August 25, 2008; revised January 26, 2009. First published April 17, 2009; current version published October 2, 2009. This work was supported in part by the U.S. National Science Foundation under Award 0424546. The review of this paper was coordinated by Prof. J. Li. B. Rong is with the Communications Research Centre Canada, Ottawa, ON K2H 8S2, Canada (e-mail: [email protected]). Y. Qian is with the National Institute of Standards and Technology, Gaithersburg, MD 20899-8920 USA (e-mail: [email protected]). K. Lu is with the Department of Electrical and Computer Engineering, University of Puerto Rico at Mayagüez, Mayagüez, PR 00681-9042 USA (e-mail: [email protected]). R. Q. Hu is with Nortel Networks, Richardson, TX 75082 USA (e-mail: [email protected]). M. Kadoch is with the Department of Electrical Engineering, Ecole de Technologie Superieure, Universite du Quebec, Montreal, QC H3C 1K3, Canada (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TVT.2009.2021062

in the WMN by taking advantage of distributed computing technology. In the literature, an extensive study has been conducted to reduce or eliminate the channel scan delay during the WMN handoff process [6]–[11]. The effort of reducing the channel scan delay basically includes two possible approaches, i.e., reducing the number of channels to scan [6], [7] and reducing the time spent on scanning each channel [8]. Some researchers have also attempted to completely eliminate handoff delay by using sensor overlay [9], multiple radios [10], and proactive scan [11]. Our work is fundamentally different from previous works, because we consider the network layer and above. Particularly, we employ an MA to save the delays incurred by reassociation, call admission control (CAC), rerouting, and resource reservation. Our design aims at benefiting the real-time applications with stringent quality-of-service (QoS) requirements. Our approach assigns each mesh client an MA, namely, a client MA, residing on the attached mesh router. If a mesh client moves to a new location and changes its mesh router, the MA migrates as well. Particularly, if the mesh client intends to make a handoff, the client MA will move to the new mesh router beforehand and presetup a backup connection for the handoff call. Then, the mesh client will accomplish the handoff process and resume the call on backup connection. As a result, our approach can significantly reduce handoff delay while achieving high computing efficiency. For the MA-based handoff, it is very important to employ the CAC mechanism in the mesh router. CAC is the process of admitting/rejecting new calls that originate in the coverage of a given mesh router or handoff calls that move into the coverage of a given mesh router while ensuring uninterrupted service of existing connections. CAC has received significant attention in the WMN for the following two reasons: First, CAC is a critical step for the provision of QoS-guaranteed services because it can prevent the system capacity from being overused. Second, CAC can give handoff calls a higher priority than new calls. In the literature, studies on CAC in traditional cellular networks can loosely be classified into two categories [12], i.e., guard channel (GC) schemes and queuing priority (QP) schemes. GC schemes reserve a number of channels in each cell for exclusive access by handoff calls [13]–[16]. It has been shown that GC schemes are easy to deploy but are not good in terms of channel efficiency. On the other hand, QP schemes accept calls whenever there are free channels. When all channels are busy, new calls are blocked, while handoff calls are queued [17], [18], or all arriving calls are queued with

0018-9545/$26.00 © 2009 IEEE

4566


Fig. 1. Example of a WMN.

certain rearrangements in the queue [19], [20]. QP schemes are not practically feasible for real-time multimedia services, if cell overlaps are not wide enough to allow handoff calls to queue and wait for a certain time for channels to become available. Based on GC and QP schemes, some improved CAC approaches were proposed to facilitate the handoff in thirdgeneration/fourth-generation (3G/4G) cellular networks, where data service is introduced to coexist with the traditional phone service. For example, Zhuang et al. [21] and Zhang et al. [22] developed CAC schemes for the scenario where the multimedia applications can adapt and trade off some of their QoS requirements at the packet level to improve the probability of handoff success. Kibria and Jamalipour [23] proposed a fair CAC algorithm that prioritizes vertical handoff processes in a multitraffic and hierarchical beyond third-generation (B3G) network. Compared with previous works on cellular networks, CAC in WMNs has some unique characteristics and challenges. First, WMN CAC mainly considers bandwidth allocation, whereas cellular CAC considers channel allocation. Second, WMN CAC has to handle heterogeneous traffic load, whereas cellular CAC only handles limited traffic genre. For the foregoing two reasons, WMN CAC has to confront a more flexible and challenging environment, as compared with its cellular counterpart. In this paper, we develop a proportional threshold structured optimal effective bandwidth (PTOEB) policy for CAC on the mesh router. This policy adopts a proportional threshold structure, gives handoff calls and new calls different priorities, and obtains a low average blocking probability. Since it is intractable to exactly locate the policy, a genetic algorithm (GA) will be utilized as the fast computational approach to achieve a near-optimal solution. Moreover, the performance of our proposed CAC scheme is evaluated by extensive analysis and simulation study.

The rest of this paper is organized as follows: We first introduce the WMN handoff challenges in Section II. We then propose an MA-based handoff architecture in Section III. In Section IV, we develop a PTOEB CAC policy, as well as the corresponding GA approximation scheme. In Section V, we evaluate the performance of our proposed CAC scheme in a WMN handoff environment through a simulation study. In the end, Section VI concludes this paper. II. H ANDOFF C HALLENGES IN WMNs A. WMNs As shown in Fig. 1, a WMN consists of two types of nodes: mesh routers and mesh clients. The mesh routers form an infrastructure of the mesh backbone for mesh clients. In general, mesh routers have minimal mobility and operate just like a network of fixed routers, except that they are connected by wireless links through wireless technologies such as IEEE 802.11. We observe from Fig. 1 that a WMN can access the Internet through a gateway mesh router, which is connected to the internet protocol (IP) core network with physical wires. In a WMN, every mesh router is equipped with a traffic aggregation device (similar to an 802.11 access point) that interacts with individual mesh clients. The mesh router relays aggregated data traffic of mesh clients to and from the IP core network. Typically, a mesh router has multiple wireless interfaces to communicate with other mesh routers, and each wireless interface corresponds to one wireless channel. These wireless channels have different characteristics, because wireless interfaces are running on different frequencies and built on either the same or different wireless access technologies, e.g., IEEE 802.11a/b/g/n. It is also possible that directional antennas are employed on some interfaces to establish wireless channels over long distances.

RONG et al.: MA-BASED HANDOFF IN WIRELESS MESH NETWORKS: ARCHITECTURE AND CAC

4567

B. Handoff Challenges Mesh clients achieve Internet access through mesh routers. A mesh client quite often moves from the coverage of one mesh router to that of another. As a result, it becomes an urgent task in WMNs to maintain the ongoing connections of roaming users. The mobile IP and related protocols can be applied to WMNs to support user mobility, but they only focus on the IP identity problem [24], [25]. In this paper, we investigate another important aspect of user mobility support, i.e., the handoff process. Ideally, WMN handoff should be accomplished with low computing cost and short latency so that the handoff process can be completely transparent to mesh clients. In this paper, we define a WMN that offers the above handoff function as a seamless handoffed WMN. Most WMNs today require specially modified clients to transfer connectivity from one mesh router to another. Although some of them give the appearance of continuous connectivity to a roaming client, handoff delay can be as long as several seconds [26]. This delay is unacceptable for real-time applications, such as voice over IP (VoIP) or videoconferencing. In the current 802.11 implementation, the handoff consists of two phases, i.e., channel scanning and connection reestablishment. During channel scanning, the mesh router scans all channels to collect the information about neighboring mesh routers. During connection reestablishment, the mesh client first registers to the new mesh client through authentication then proceeds to the postregistration stage, which includes reassociation, CAC, rerouting, and resource reservation to meet the requirements of real-time applications. To reduce handoff delay, previous studies mainly focused on shortening the channel scan latency [6]–[11]. Different from previous works, in this paper, we propose an MA-based handoff architecture, where an MA takes care of the handoff signaling process in the network layer and above. Specifically, an MA accomplishes the tasks of the postregistration stage, such as reassociation, CAC, rerouting, and resource reservation, prior to the actual handoff. As a result, the handoffed mesh client is able to immediately continue connectivity after registering to the new mesh router. III. MA-B ASED H ANDOFF A RCHITECTURE IN WMNs In this section, we propose an MA-based handoff architecture, which offers seamless and fast handoff to support VoIP and other real-time applications. In our approach, all the handoff logics are done by the MA, and only the standard medium-access control protocol and IP are used. Therefore, it is compatible with any 802.11 mobile device, regardless of the vendor or architecture. A. Introduction to an MA An MA is an executing program that can migrate during execution from machine to machine in a heterogeneous network. In other words, the agent can suspend its execution, migrate to another machine, and then resume execution on the new machine from the point at which it left off. On each machine,

Fig. 2. Architecture of MA-based handoff in WMNs.

the agent interacts with local resources to accomplish its task. There are a number of MA platforms currently existing, for example, [27]–[29]. MAs have several advantages in developing distributed computing applications. By migrating to the information resource, an agent can locally invoke resource operations, eliminating the network transfer of intermediate data. By migrating to the other side of an unreliable network link, an agent can continue executing, even if the network link goes down, making MAs particularly attractive in mobile computing environments. By choosing different migration strategies, an agent can adapt itself to different tasks and network conditions, achieving full flexibility and customization. It is appropriate to deploy MAs in a WMN, since a WMN is a typical distributed system with the feature of “mobility.”

B. MA-Based Handoff To provide seamless handoff, we apply MA technology to WMNs. As shown in Fig. 2, in our solution, each mesh client is assigned a “client MA.” The mesh client places its client MA in the mesh router that it registers with. If the mesh client moves from the coverage of one mesh router to that of another mesh router, the client MA also migrates. Our approach can work well as a supplement with the channel scan schemes proposed in a previous study. As an example, we demonstrate in the following the integration of our approach with the proactive scan scheme proposed in [11]. We study the scenario that a mesh client moves from the coverage of one mesh router to that of another mesh router during a call. To eliminate the overall handoff latency, we can employ a proactive scan scheme to counteract channel scan delay and our MA approach to counteract connection reestablishment delay. Particularly, when the scan trigger of a proactive scan scheme is fired, the mesh client will actively probe channels and choose the appropriate neighboring mesh router for handoff. Then, the client MA will move from the current mesh router to the chosen mesh router and complete the processes of reassociation, CAC, rerouting, resource reservation, etc. Once the handoff trigger is fired, the mesh client will register to the new mesh router and resume all the connections using the facilities that have been prepared by the client MA earlier.

4568


calls in the CAC process, since users are much more sensitive to call dropping than to call blocking. In the rest of this section, we design a CAC policy, named PTOEB, which adopts a proportional threshold structure to implement CAC, gives handoff calls and new calls different priorities, and offers a high effective bandwidth. A. System Model

Fig. 3. Process of MA-based handoff.

Fig. 3 demonstrates that there are five steps in the joint handoff process of the proactive scan scheme and our MA approach. First, the scan trigger of the proactive scan scheme activates the channel scan, which locates the new mesh router for handoff. Second, the mesh client will inform its client MA on the current mesh router which one is the new mesh router. Third, the current mesh router transfers the client MA to the new mesh router in the neighborhood, and the client MA will presetup backup connections on the new mesh router to prepare for seamless handoff. The presetup of backup connections usually involves reassociation for context switching between the old access point (AP) and the new AP by the interaccess point protocol [30], interaction with the CAC module for resource reservation, and negotiation with the routing protocol for network layer path reestablishment. Fourth, once the backup connection is built up, the client MA will notify the mesh client that it is ready for handoff. Finally, the mesh client receives the notification and waits for the fire of the handoff trigger to register to the new mesh router and complete the handoff. The foregoing illustrations show that before the actual handoff occurs in the fifth step, the client MA has already constructed a backup connection on the new mesh router in the third step. As a result, overall handoff delay only involves registration delay, which is spent on the authentication information exchange between the mesh client and the new mesh router. In addition to reducing handoff delay, MA-based handoff can also achieve high computing efficiency. The client MA executes handoff logics on the mesh router where the network computing resource is affluent and, thus, release the burden on the mesh client, which is dedicated to running user applications. IV. CAC S CHEME FOR WMN H ANDOFF It is a challenging task to provide mobile users with QoSguaranteed services, particularly when the traffic load in WMNs is heavy. Many researchers believe that CAC plays a critical role in overcoming this challenge. Traditionally, CAC aims to maximize the number of admitted sessions while meeting their QoS requirements. However, in a WMN handoff environment, we also have to consider the issue of differentiated priorities. That is, handoff calls have to be given more preference than new

A system model of CAC for WMN handoff can be formulated as follows: We assume that there are M classes of traffic loads in the network. On a given mesh router, all the traffic loads share the overall B units of physical access bandwidth, and each class of traffic load consists of both new calls and handoff calls. With regard to the class i traffic, we assume the following. 1) The requests arrive from a random process with an averhf age rate λnw i for new calls and λi for handoff calls. 2) The average connection holding time is 1/μnw i s for new s for handoff calls. calls and 1/μhf i 3) The bandwidth requirement of a connection are fixed to hf nw and bhf bnw i = bi = bi , where bi i represent the bandwidth requirements of class i new call and handoff call, respectively. Then, CAC on the mesh router is responsible to accept or reject connection requests based on the system state, the type of connections, and the QoS requirements of connections. Let the bandwidth requirement vector be b = (b1 , b2 , . . . , bM , bM +1 , bM +2 , . . . , b2M ) nw nw hf hf hf = bnw 1 , b2 , . . . , bM , b1 , b2 , . . . , bM

(1)

the traffic intensity vector be ρ = (ρ1 , ρ2 , . . . , ρM , ρM +1 , ρM +2 , . . . , ρ2M ) λ1 λ 2 λM λM +1 λM +2 λ2M , ,..., , , ,..., = μ1 μ2 μM μM +1 μM +2 μ2M nw nw hf hf hf = ρ1 , ρ2 , . . . , ρnw M , ρ1 , ρ2 , . . . , ρM nw nw hf λ1 λ2 λnw λhf λhf 2 M λ1 M = , , . . . , , , , . . . , (2) hf μhf μnw μnw μnw μhf 1 2 M μ1 2 M and the system state vector be n = (n1 , n2 , . . . , nM , nM +1 , nM +2 , . . . , n2M ) nw nw hf hf hf = nnw 1 , n2 , . . . , nM , n1 , n2 , . . . , nM

(3)

where nnw and nhf i i are the numbers of class i new calls and handoff calls running on the mesh router, respectively. We then define the ith traffic load (1 ≤ i ≤ 2M ) as the traffic load characterized by the parameter set (bi , ρi , ni ). Clearly, the ith traffic load belongs to new call traffic if 1 ≤ i ≤ M , and it belongs to handoff traffic if M + 1 ≤ i ≤ 2M . Moreover, both the ith and the M + ith traffic loads are categorized into the class i traffic. Based on foregoing discussions, we can further define ΩCS as the set of all possible system states, which can be expressed as ΩCS = {n|n · b ≤ B}. Under this definition, the subscript CS


4569

stands for “complete sharing,” which means that an incoming connection will be accepted if sufficient bandwidth resources are available in the system. We can now define a CAC policy, denoted by Ω, as an arbitrary subset of ΩCS . Given Ω, a connection request will be accepted if and only if the system state vector remains in Ω after the connection being accepted. B. PTOEB CAC Policy 1) Requirement of a Differentiated Priority: Other than CAC in high-speed wired networks [31]–[33], CAC becomes much more complicated in wireless networks because of user mobility. An accepted call that has not completed in the coverage of the current mesh router may have to be handed off to another mesh router. During the process, the call may not be able to gain a session channel in the new mesh router to continue its service due to the limited resource and, thus, have to be dropped. Correspondingly, the new calls and handoff calls have to be differently treated in terms of resource allocation. That is, handoff calls are usually assigned a higher priority over new calls, since users are much more sensitive to call dropping than to call blocking. 2) Requirement of a Statistical Effective Bandwidth: In general, mesh clients want to maximize the network throughput and have Internet access of the broadest bandwidth. Therefore, they prefer a CAC policy that produces the maximal statistical effective bandwidth. For a given CAC policy Ω, we define the statistical effective bandwidth that it can achieve as (n · b)PΩ (n) (4) BE (Ω) = n∈Ω

where PΩ (n) is the steady-state probability that the system is in state n. Based on (4), we further define the normalized statistical effective bandwidth of the CAC policy Ω as N (Ω) = BE (Ω)/B. BE

(5)

If a policy Ω satisfies the coordinate convex condition and the arrival and service processes are both memoryless, then PΩ (n) can be calculated by (as shown in [34]) 2M

PΩ (n) =

1 ρni i , G(Ω) i=1 ni !

n ∈ Ω

(6)

where G(Ω) =

2M

ρni i /ni !.

(7)

n∈Ω i=1

It is worth noting that the memoryless property of the service hf process implies that μnw i = μi in our CAC model. Moreover, the blocking probability of the ith traffic load is G Ωbi P bi (Ω) = , 1 ≤ i ≤ 2M (8) G(Ω) where Ωbi = {n|n ∈ Ω & n + ei ∈ / Ω}, and ei is a 2M dimensional vector of all 0s, except its ith element, which is 1.

Fig. 4. Two-dimensional threshold structured CAC policy.

To further reveal the relationship between BE (Ω) and P bi (Ω), we have the following Lemma. Lemma 1: For any CAC policy that satisfies the coordinate convex condition, the statistical effective bandwidth defined in (4) can be calculated by BE =

2M i=1

bi ρi −

2M

bi ρi P bi (Ω).

(9)

i=1

2M In (9), i=1 bi ρi stands for the overall arriving traffic 2M load, and i=1 bi ρi P bi (Ω) stands for the blocked traffic load. Clearly, BE (Ω) is equal to the bandwidth of all accepted traffic, which can be obtained by subtracting the blocked traffic load from the overall arriving traffic load. Therefore, Lemma 1 holds for any policy that satisfies the coordinate convex condition. 3) Tradeoff Between Differentiated Priorities and the Statistical Effective Bandwidth: As stated earlier in this paper, handoff requires differentiated priorities, whereas mesh clients require the maximal statistical effective bandwidth. To balance the aforementioned two requirements, we propose a CAC policy of PTOEB, which is derived from the traditional threshold structured CAC. When choosing the structure of the CAC policy, we mainly consider two criteria, i.e., the implementing complexity and the efficiency. In particular, if the optimal solution for the multidimensional CAC problem has complicated geometry, we consider that it might not be realistic for practical implementation. A threshold structured CAC policy can be defined as

(10) ΩTH = n ∈ ΩCS |ni < Nith where 0 ≤ Nith ≤ B/bi is the threshold imposed on the ith traffic load. Fig. 4 illustrates a threshold structured CAC policy for two traffic loads. It is shown that the 2-D threshold structured CAC policy is determined by three boundaries, i.e., the threshold on n1 , the threshold on n2 , and the CS boundary. In terms of the implementing complexity, the threshold structured CAC policy has an obvious advantage due to its simple geometry. In terms of efficiency, Ross and Tsang [31] stated that the threshold structured CAC policy has the best performance in the case of two dimensions, but the optimal solution of arbitrary multiple dimensions remains an open problem. However, as

4570


Fig. 5. Two-dimensional CAC policy of the proportional threshold structure.

shown in the literature [35], [36], the threshold structured CAC policy has widely been accepted in practice as a highly efficient CAC structure for any multidimensional scenarios. The PTOEB CAC policy adopts a proportional threshold structure, which combines a threshold structure with a proportional constraint. To construct a proportional threshold struc th = tured CAC policy, we first apply the threshold vector N th th th th , NM , N , . . . , N ) as the upper (N1th , N2th , . . . , NM +1 M +2 2M th bound to the system state vector n so that ni ≤ Ni (i = 1, 2, . . . , M, M + 1, . . . , 2M ). We then introduce the following proportional constraint to give handoff calls more priority than new calls: ρi th th N , B/bi , Ni = min x (i = 1, 2, . . . , M ) (11) ρM +i M +i where 0 ≤ x ≤ 1 is the proportional factor. Inside min{x(ρi / th th ρM +i )NM +i , B/bi }, the first term x(ρi /ρM +i )NM +i represents the proportional relationship between the threshold on new calls and the threshold on handoff calls, whereas the second term B/bi represents the capacity limit. To clearly explain the foregoing concept, Fig. 5 illustrates a simple example of a 2-D CAC policy, which handles one new call traffic with the parameter (n1 , N1th ) and one handoff call traffic with the parameter (n2 , N2th ). In this example, we th mainly address the constraint of Nith = x(ρi /ρM +i )NM +i and th thus suspend the capacity limit Ni ≤ B/bi . We define the threshold rectangle as the rectangle determined by the point set {(0, 0), (0, N2th ), (N1th , N2th ), (N1th , 0)} and the diagonal line of the threshold rectangle as the line between (0, 0) and (N1th , N2th ). Then, the slope of the diagonal line is given by Sd =

N2th ρ2 = . xρ1 N1th

(12)

Equation (12) indicates that, when the traffic intensity (ρ1 , ρ2 ) is known, the proportional factor x formulates a set of threshold structured CAC policies, which have the same diagonal line slope. The foregoing discussion can easily be extended to the proportional threshold structured CAC policy of any dimensions. Once x is determined by the network administrator, it formulates a set of proportional threshold structured policies

that are subject to the proportional factor x. Among them, the CAC policy of PTOEB is the policy that produces the maximal statistical effective bandwidth. 4) CAC Analysis From the Perspective of the Blocking Probability: The blocking probability is often employed to evaluate the performance of a CAC policy, since it can directly reflect the user feelings on the service. In the following, we analyze the PTOEB policy from the perspective of the blocking probability. When designing CAC, we address the issue that, inside one unique traffic class, handoff calls should have a higher priority over new calls. In a real network, besides the priority ratio of handoff calls, there are also many other factors that influence the blocking probability of a single traffic class, including system capacity, traffic intensities, and QoS requirements of all classes, multiple optimization objectives of the CAC policy, etc. Therefore, the blocking probability of a single traffic class does not directly and clearly carry the information of the relationship between new calls and handoff calls. In this respect, the average blocking probabilities of new calls and handoff calls become a better way to reflect our concern. We define the average blocking probability of new calls, handoff calls, and all calls as follows. The average blocking probability of new calls is M bi ρi P bi (Ω) nw . (13) P bAvg (Ω) = i=1 M i=1 bi ρi The average blocking probability of handoff calls is 2M i=M +1 bi ρi P bi (Ω) P bhf (Ω) = . 2M Avg i=M +1 bi ρi The average blocking probability of all calls is 2M bi ρi P bi (Ω) P bAvg (Ω) = i=1 . 2M i=1 bi ρi

(14)

(15)

From Lemma 1, we can derive that BE P bAvg (Ω) = 1 − 2M . i=1 bi ρi

(16)

Equation (16) shows that P bAvg (Ω) is directly related to the statistical effective bandwidth BE (Ω). Considering that 2M i=1 bi ρi is a constant when the traffic load is known, P bAvg (Ω) decreases as the statistical effective bandwidth BE (Ω) increases. We further define the priority ratio of handoff calls as Rpr (Ω) =

hf P bnw Avg (Ω) − P bAvg (Ω) P bAvg (Ω)

(17)

which reflects the preference that handoff calls can obtain during the call admission process. If CS is employed as the CAC policy, handoff calls and new calls are equally treated. Therefore, the CS policy yields of a result of no priority, which can formally be represented by Rpr (Ω) = 0. Using CS as reference, when Rpr (Ω) > 0, we conclude that handoff calls have a higher priority than new calls; when Rpr (Ω) < 0,


4571

we conclude that handoff calls have a lower priority than new calls. On one hand, the requirement of a high statistical effective bandwidth expects the CAC policy to produce a low value of P bAvg (Ω). On the other hand, the requirement of a differentiated priority expects that the value of Rpr (Ω) could be as high as possible. To meet both requirements, the PTOEB CAC policy adjusts the value of the proportional factor x to achieve the desired blocking probabilities.

C. GA for Near-Optimal Solutions A proportional threshold structured CAC policy can be determined by and only by the second half of the threshold th , i.e., N hf = (N th , N th , . . . , N th ), since the vector N M +1 M +2 2M th th , i.e., N nw = (N th , N th , . . . , N th ), can be first half of N 1 2 M th derived from the second half by the proportional threshold constraint. In this respect, the set of all possible proportional threshold structured CAC policies can be described as a space hf . As a result, the task of finding of all possible vectors of N th the PTOEB CAC policy can be modeled as an optimization hf to achieve problem, whose goal is to optimize the value of N th the optimal statistical effective bandwidth. To solve this optimization problem, a straightforward method is to employ bruteforce search. Although the method of brute-force search can obtain the exact optimal solution, it usually has tremendous computational complexity. Correspondingly, we employ a GA to search for the near-optimal solution [37]. A GA is an adaptive heuristic search program that applies the principles of evolution found in nature. The GA combines selection, crossover, and mutation operators with the goal of finding the solution of best fitness to a problem. Here, fitness is a special GA term, which refers to the objective function of the optimization problem. In our optimization problem, fitness is defined as the statistical effective bandwidth function in (4). In a GA, the solution to a problem is called a chromosome. A chromosome is made up of a collection of genes, which are simply the parameters to be optimized. A GA creates an initial population with a collection of chromosomes, evaluates this population, and evolves the population through multiple generations using the genetic operators in the search for a good solution of the optimization problem, until a specified termination criterion is met. For the CAC policies of the proportional threshold structure, the searching space can be represented by a vec hf = (N th , N th , . . . , N th ), which tor of M variables N M +1 M +2 2M th is subject to the condition Nith ≤ B/bi . The vector of th th th (NM +1 , NM +2 , . . . , N2M ) also serves as the chromosome in the GA. To solve the statistical effective bandwidth optimization problem, we define the genetic operators in the list that follows. 1) Selection operator. The “Roulette” approach is chosen as the selection operator for statistical effective bandwidth optimization. In “Roulette,” the chance of a chromosome getting selected is proportional to its fitness. This is where the principle of survival of the fittest comes into play.

Fig. 6. Crossover operator in the GA for CAC optimization.

Fig. 7. Mutation operator in the GA for CAC optimization.

2) Crossover operator. “One Point Crossover” is employed in statistical effective bandwidth optimization. “One Point Crossover” randomly selects a crossover point within a chromosome and then interchanges the two parent chromosomes at this point to produce two new offspring. Consider the following two parents: th th th Parent a = (NM +1 (a), NM +2 (a), . . . , N2M (a)). th th th Parent b = (NM +1 (b), NM +2 (b), . . . , N2M (b)). Fig. 6 shows the detailed function of the crossover operator in the GA for CAC optimization. 3) Mutation operator. We utilize “Gaussian Mutation” for statistical effective bandwidth optimization. “Gaussian Mutation” adds a unit Gaussian distributed random value to the chosen gene. The new gene value is clipped if it falls outside the user-specified lower or upper bound for that gene. To make it clear, an example is given in Fig. 7. 4) Termination method. We use “Fitness Convergence” as the termination criterion in our application. This termination criterion stops the evolution when fitness is deemed converged.

4572


TABLE I TRAFFIC LOAD CONFIGURATION IN THE COVERAGE OF A MESH ROUTER

Fig. 8. Normalized statistical effective bandwidth while varying the proportional factor x.

V. P ERFORMANCE E VALUATION

Fig. 9. Average blocking probability of new calls, handoff calls, and all calls while varying the proportional factor x.

In this section, numerical results are presented to demonstrate the performance of our proposed CAC scheme in a WMN handoff environment. We first evaluate the performance of the PTOEB CAC policy in a single mesh-router cell. We then conduct a performance comparison among our CAC scheme and other existing CAC schemes. A. Performance of the PTOEB CAC Policy In our simulation, we emulate the environment of a multimedia WMN, which is dominated by real-time audio/video applications. We consider that WMN users may access a variety of services, such as VoIP, IP television, and video surveillance [38]–[43]. These services further employ different video compression standards, including MPEG-2, MPEG-4, etc. [44], [45]. Particularly, we assume that five classes of traffic share a total of 360 Mb/s physical access bandwidth in the coverage of a mesh router, and each traffic class includes both new calls and handoff calls, as shown in Table I. Using the GA discussed in Section IV-C, we demonstrate the performance of the PTOEB CAC policy in Figs. 8–10 while varying the proportional factor x from 0% to 100%. Specifically, the normalized statistical effective bandwidth in Fig. 8, the blocking probability in Fig. 9, and the priority ratio of handoff calls in Fig. 10 are achieved from near-optimal solutions when the GA reaches convergence. In our CAC policy of PTOEB, we have taken into account two criteria, i.e., priority differentiation and statistical effective bandwidth. Here, priority differentiation highlights the user feeling on call dropping, whereas statistical effective bandwidth

Fig. 10.

Priority ratio of handoff calls while varying the proportional factor x.

highlights the user feeling on call blocking. Figs. 8–10 illustrate that, as the proportional factor x increases from 0% to 100%, the effective bandwidth goes up, and the blocking probability of all calls and the priority ratio of handoff calls go down. In other words, small x results in a high probability of call blocking, whereas large x results in a high probability of call dropping. For example, when x = 0%, PTOEB gives handoff calls overwhelming preference over new calls, yielding a solution of good differentiated priorities but the lowest statistical effective bandwidth and the highest blocking probability of all calls. In fact, x = 0% implies that the system only accepts handoff calls. By contrast, when x = 100%, PTOEB gives new calls and handoff calls the same priority, yielding a solution


4573

Fig. 11. Average normalized statistical effective bandwidth per generation versus iteration steps. Fig. 12.

Normalized statistical effective bandwidth of different CAC schemes.

of the highest statistical effective bandwidth and the lowest blocking probability of all calls but no differentiated priority. Therefore, to achieve good performance from PTOEB, we need to find the value of x that could well balance the user feelings on call dropping and call blocking. In this paper, we suggest x = 50% as a generally appropriate configuration to most WMNs. Next, we demonstrate the convergence speed of our GA (x = 50%) in Fig. 11, using the average fitness (statistical effective bandwidth) per generation as the evaluation criterion. The results show that our GA can reach convergence after 20 generations; thus, it can guarantee efficient implementation.

B. Performance Comparison Among Different CAC Policies We compare our CAC scheme with CS and the rigiddivision-based GC priority (RGS) scheme proposed in [16]. In the RGS scheme, there is a rigid division between the total channels of a cell. The channels are divided into two different groups, called the common channel group and the GC group. The channels in the common channel group can be used by new calls and handoff calls. The channels in the GC group can only be used by handoff calls. Using the handoff model of WMNs in this paper, the RGS scheme can be reformulated by the following two restrictions:

nnw ·bnw ≤ B −BG , restriction of guard channel (18) n ·b ≤ B, restriction of capacity limitation

nw nw nw nw nw where nnw = {nnw 1 , n2 , . . . , nM }, bnw = {b1 , b2 , . . . , bM }, and BG is the number of bandwidths for the GC group. Equation (18) shows that the RGS scheme has only one adjustable parameter BG , which leads to a simpler CAC structure than our proposed scheme. We present the numerical results in Figs. 12 and 13 to demonstrate the statistical effective bandwidth and handoff-call priority ratio of PTOEB, RGS, and CS, respectively. In this simulation scenario, we vary the physical access bandwidth of the mesh router from 300 to 420 Mb/s while keeping the other

Fig. 13. Priority ratio of handoff calls for different CAC schemes.

parameters the same as in the previous simulation. We configure x = 50% for PTOEB and BG = 25%B for RGS. Figs. 12 and 13 show that CS obtains a high statistical effective bandwidth but no priority for handoff calls, RGS obtains an excellent handoff call priority ratio but a low statistical effective bandwidth, and our proposed PTOEB CAC scheme can simultaneously achieve a high statistical effective bandwidth and a high priority ratio of handoff calls.

VI. C ONCLUSION To accommodate the mobility feature in WMNs, it is important to provide an efficient seamless handoff solution. In this paper, we have presented an MA-based handoff architecture, which can achieve short handoff delay and high computing efficiency at the same time. To give preference to handoff calls on mesh routers, we further developed a PTOEB CAC policy. Since it is intractable to exactly locate the aforementioned optimal CAC policy, we used a GA to obtain a near-optimal solution. Numerical results show that our CAC scheme can give handoff calls and new calls differentiated priorities while achieving a high statistical effective bandwidth.

4574


R EFERENCES [1] I. F. Akyildiz and X. Wang, “A survey on wireless mesh networks,” IEEE Commun. Mag., vol. 43, no. 9, pp. S23–S30, Sep. 2005. [2] A. Raniwala and T. Chiueh, “Architecture and algorithms for an IEEE 802.11-based multi-channel wireless mesh network,” in Proc. IEEE INFOCOM, Mar. 2005, vol. 3, pp. 2223–2234. [3] H. Jiang, W. Zhuang, X. Shen, A. Abdrabou, and P. Wang, “Differentiated services for wireless mesh backbone,” IEEE Commun. Mag., vol. 44, no. 7, pp. 113–119, Jul. 2006. [4] R. Huang, C. Zhang, and Y. Fang, “A mobility management scheme for wireless mesh networks,” in Proc. IEEE GLOBECOM, Nov. 2007, pp. 5092–5096. [5] H. Wang, Q. Huang, Y. Xia, Y. Wu, and Y. Yuan, “A network-based local mobility management scheme for wireless mesh networks,” in Proc. IEEE WCNC, Mar. 2007, pp. 3792–3797. [6] A. Mishra, M. Shin, and W. A. Arbaush, “Context caching using neighbor graphs for fast handoffs in a wireless network,” in Proc. IEEE INFOCOM, Mar. 2004, vol. 1, pp. 351–361. [7] S. Pack, H. Jung, T. Kwon, and Y. Choi, “A selective neighbor caching scheme for fast handoff in IEEE 802.11 wireless networks,” in Proc. IEEE ICC, May 2005, vol. 5, pp. 3599–3603. [8] I. Ramani and S. Savage, “SyncScan: Practical fast handoff for 802.11 infrastructure networks,” in Proc. IEEE INFOCOM, Mar. 2005, vol. 1, pp. 675–684. [9] S. Waharte, K. Ritzenthaler, and R. Boutaba, “Selective active scanning for fast handoff in WLAN using sensor networks,” in Proc. 6th IFIP/IEEE Int. Conf. MWCN, Oct. 2004, pp. 59–70. [10] V. Brik, A. Mishra, and S. Banerjee, “Eliminating handoff latencies in 802.11 WLANs using multiple radios: Applications, experience, and evaluation,” in Proc. ACM/USENIX IMC, Oct. 2005, pp. 299–304. [11] H. Wu, K. Tan, Y. Zhang, and Q. Zhang, “Proactive scan: Fast handoff with smart triggers for 802.11 wireless LAN,” in Proc. IEEE INFOCOM, May 2007, pp. 749–757. [12] Y. Fang and Y. Zhang, “Call admission control schemes and performance analysis in wireless mobile networks,” IEEE Trans. Veh. Technol., vol. 51, no. 2, pp. 371–382, Mar. 2002. [13] D. Hong and S. S. Rappaport, “Traffic model and performance analysis for cellular radio telephone systems with prioritized and nonprioritized handoff procedure,” IEEE Trans. Veh. Technol., vol. VT-35, no. 3, pp. 77–92, Aug. 1986. [14] Y. B. Lin, S. Mohan, and A. Noerpel, “Queueing priority channel assignment strategies for handoff and initial access for a PCS network,” IEEE Trans. Veh. Technol., vol. 43, no. 3, pp. 704–712, Aug. 1994. [15] R. Ramjee, D. Towsley, and R. Nagarajan, “On optimal call admission control in cellular networks,” Wirel. Netw., vol. 3, no. 1, pp. 29–41, Mar. 1997. [16] M. D. Kulavaratharasah and A. H. Aghvami, “Teletraffic performance evaluation of microcellular personal communication networks (PCNs) with prioritized handoff procedures,” IEEE Trans. Veh. Technol., vol. 48, no. 1, pp. 137–152, Jan. 1999. [17] E. D. Re, R. Fantacci, and G. Giambene, “Handover queueing strategies with dynamic and fixed channel allocation techniques in low earth orbit mobile satellite systems,” IEEE Trans. Commun., vol. 47, no. 1, pp. 89–102, Jan. 1999. [18] S.-P. Chung and M.-T. Li, “Prioritized call admission control of multiservice hierarchical wireless network,” IEICE Trans. Commun., vol. E88-B, no. 6, pp. 2566–2577, Jun. 2005. [19] C. Chang, C. J. Chang, and K. R. Lo, “Analysis of a hierarchical cellular system with reneging and dropping for waiting new calls and handoff calls,” IEEE Trans. Veh. Technol., vol. 48, no. 4, pp. 1080–1091, Jul. 1999. [20] V. K. N. Lau and S. V. Maric, “Mobility of queued call requests of a new call-queuing technique for cellular systems,” IEEE Trans. Veh. Technol., vol. 47, no. 2, pp. 480–488, May 1998. [21] W. Zhuang, B. Bensaou, and K. C. Chua, “Adaptive quality of service handoff priority scheme for mobile multimedia networks,” IEEE Trans. Veh. Technol., vol. 49, no. 2, pp. 494–505, Mar. 2000. [22] J. Zhang, J. W. Mark, and X. Shen, “An adaptive handoff priority scheme for wireless MC-CDMA cellular networks supporting multimedia applications,” in Proc. IEEE GLOBECOM, Nov. 2004, vol. 5, pp. 3088–3092. [23] M. R. Kibria and A. Jamalipour, “NXG04-5: Fair call admission control for prioritizing vertical handoff in multi-traffic B3G networks,” in Proc. IEEE GLOBECOM, Nov. 2006, pp. 1–5. [24] R. L. Geiger, J. D. Solomon, and K. J. Crisler, “Wireless network extension using mobile IP,” IEEE Micro, vol. 17, no. 6, pp. 63–68, Nov./Dec. 1997. [25] S. Mohanty and I. F. Akyildiz, “Performance analysis of handoff techniques based on mobile IP, TCP-migrate, and SIP,” IEEE Trans. Mobile Comput., vol. 6, no. 7, pp. 731–747, Jul. 2007.

[26] H. Velayos and G. Karlsson, “Techniques to reduce IEEE 802.11b MAC layer handover time,” Royal Inst. Technol. (KTH), Stockholm, Sweden, KTH Tech. Rep. TRITA-IMIT-LCN R 03:02, Apr. 2003. ISRN KTH/IMIT/LCN/R-03/02-SE. [27] G. Karjoth, D. B. Lange, and M. Oshima, “A security model for Aglets,” IEEE Internet Comput., vol. 1, no. 4, pp. 68–77, Jul./Aug. 1997. [28] J. Kiniry and D. Zimmerman, “A hands-on look at Java mobile agents,” IEEE Internet Comput., vol. 1, no. 4, pp. 21–30, Jul./Aug. 1997. [29] L. M. Silva, G. Soares, P. Martins, V. Batista, and L. Santos, “The performance of mobile agent platforms,” in Proc. 1st Int. Symp. Agent Syst. Appl./3rd Int. Symp. Mobile Agents, Oct. 1999, pp. 270–271. [30] IEEE Trial-Use Recommended Practice for Multi-Vendor Access Point Interoperability via an Inter-Access Point Protocol Across Distribution Systems Supporting IEEE 802.11 Operation, IEEE Std. 802.11F-2003, 2003. [31] K. W. Ross and D. H. K. Tsang, “The stochastic knapsack problem,” IEEE Trans. Commun., vol. 37, no. 7, pp. 884–895, Jul. 1989. [32] E. Altman, T. Jimenez, and G. Koole, “On optimal call admission control in resource-sharing system,” IEEE Trans. Commun., vol. 49, no. 9, pp. 1659–1668, Sep. 2001. [33] V. Sarangan, D. Ghosh, N. Gautam, and R. Acharya, “Steady state distribution for stochastic knapsack with bursty arrivals,” IEEE Commun. Lett., vol. 9, no. 2, pp. 187–189, Feb. 2005. [34] J. S. Kaufman, “Blocking in a shared resource environment,” IEEE Trans. Commun., vol. COM-29, no. 10, pp. 1474–1481, Oct. 1981. [35] A. Gavious and Z. Rosberg, “A restricted complete sharing policy for a stochastic knapsack problem in B-ISDN,” IEEE Trans. Commun., vol. 42, no. 7, pp. 2375–2379, Jul. 1994. [36] C. C. Beard and V. S. Frost, “Prioritized resource allocation for stressed networks,” IEEE/ACM Trans. Netw., vol. 9, no. 5, pp. 618–633, Oct. 2001. [37] M. Mitchell, An Introduction to Genetic Algorithms. Cambridge, MA: MIT Press, 1996. [38] S. Ganguly, V. Navda, K. Kim, A. Kashyap, D. Niculescu, R. Izmailov, S. Hong, and S. R. Das, “Performance optimizations for deploying VoIP services in mesh networks,” IEEE J. Sel. Areas Commun., vol. 24, no. 11, pp. 2147–2158, Nov. 2006. [39] B. Rong, Y. Qian, and H.-H. Chen, “An enhanced SIP proxy server for wireless VoIP in wireless mesh networks,” IEEE Commun. Mag., vol. 46, no. 1, pp. 108–113, Jan. 2008. [40] S. Ramprashad, D. Li, U. Kozat, and C. Pepin, “An analysis of joint aggregation, bursting, routing, and rate adaptation for increasing VoIP capacity in multi-hop 802.11 networks,” IEEE Trans. Wireless Commun., vol. 7, no. 8, pp. 3128–3139, Aug. 2008. [41] I. Djama and T. Ahmed, “Cross-layer interworking of DVB-T and WLAN for mobile IPTV service delivery,” IEEE Trans. Broadcast., vol. 53, no. 1, pp. 382–390, Mar. 2007. [42] E. Shihab, L. Cai, F. Wan, A. Gulliver, and N. Tin, “Wireless mesh networks for in-home IPTV distribution,” IEEE Netw., vol. 22, no. 1, pp. 52–57, Jan./Feb. 2008. [43] W.-T. Chen, P.-Y. Chen, W.-S. Lee, and C.-F. Huang, “Design and implementation of a real time video surveillance system with wireless sensor networks,” in Proc. IEEE VTC—Spring, May 2008, pp. 218–222. [44] L. Galluccio, F. Licandro, G. Morabito, and G. Schembra, “An analytical framework for the design of intelligent algorithms for adaptiverate MPEG video encoding in next-generation time-varying wireless networks,” IEEE J. Sel. Areas Commun., vol. 23, no. 2, pp. 369–384, Feb. 2005. [45] L. Galluccio, G. Morabito, and G. Schembra, “Transmission of adaptive MPEG video over time-varying wireless channels: Modeling and performance evaluation,” IEEE Trans. Wireless Commun., vol. 4, no. 6, pp. 2777–2788, Nov. 2005.

Bo Rong (M’07) received the B.S. degree from Shandong University, Jinan, China, in 1993, the M.S. degree from Beijing University of Aeronautics and Astronautics, Beijing, China, in 1997, and the Ph.D. degree from Beijing University of Posts and Telecommunications, in 2001. He is currently a Research Scientist with the Communications Research Centre Canada, Ottawa, ON, Canada. He is also an Adjunct Professor with the Ecole de Technologie Superieure, Universite du Quebec, Montreal, QC, Canada. His research interests include modeling, simulation, and performance analysis of next-generation wireless networks.


4575

Yi Qian (M’95–SM’07) received the Ph.D. degree in electrical engineering with a concentration in telecommunication networks from Clemson University, Clemson, SC. He is currently with the National Institute of Standards and Technology, Gaithersburg, MD. He was an Assistant Professor with the Department of Electrical and Computer Engineering, University of Puerto Rico at Mayagüez (UPRM), between July 2003 and July 2007. At UPRM, he taught courses on wireless networks, network design, network management, and network performance analysis. His research and curriculum development efforts were funded by, among others, the National Science Foundation, General Motors, IBM, and PRIDCO, with more than $2 million in total awards during his four years at UPRM. Prior to joining UPRM in July 2003, he worked for several start-up companies and consulting firms, in the areas of voice over internet protocol, fiber optical switching, Internet packet video, network optimizations, and network planning as a Technical Advisor and a Senior Consultant. He also worked several years for the Wireless Systems Engineering Department, Nortel Networks, Richardson, TX, as a Senior Member of Scientific Staff and as a Technical Advisor. While at Nortel, he was a Project Leader for various wireless and satellite network product design projects, customer consulting projects, and advanced-technology research projects. He was also in charge of a wireless standard development and evaluation project with Nortel. He has publications and patents in all these areas. His current research interests include information assurance, network security, network management, network design, network modeling, simulation and performance analysis for nextgeneration wireless networks, wireless sensor networks, broadband satellite networks, optical networks, high-speed networks, and the Internet. Dr. Qian is a member of the Association for Computing Machinery.

Rose Qingyang Hu (S’95–M’98–SM’06) received the B.S.E.E. degree from the University of Science and Technology of China, Hefei, China, the M.S. degree from Polytechnic University, Brooklyn, NY, and the Ph.D. degree in electrical engineering from the University of Kansas, Lawrence. After receiving the Ph.D. degree, she was a Senior Systems Engineer for four years with Nortel Networks, Richardson, TX, and Yotta Networks. From January 2002 to June 2004, she was an Assistant Professor with the Department of Electrical and Computer Engineering, Mississippi State University. She is currently a Technical Manager with the Wireless Standards and Architecture Team, Nortel Networks, where she is responsible for Nortel’s 4G wireless technology performance evaluation and standards development. She has published about 40 journal and conference papers. She is the holder of six U.S. patents in her research areas. Dr. Hu is a member of Phi Kappa Phi and Epsilon Pi Epsilon.

Kejie Lu (S’01–M’04–SM’07) received the B.S. and M.S. degrees in telecommunications engineering from Beijing University of Posts and Telecommunications, Beijing, China, in 1994 and 1997, respectively, and the Ph.D. degree in electrical engineering from the University of Texas at Dallas, Richardson, in 2003. In 2004 and 2005, he was a Postdoctoral Research Associate with the Department of Electrical and Computer Engineering, University of Florida, Gainesville. Since 2005, he has been an Assistant Professor with the Department of Electrical and Computer Engineering, University of Puerto Rico at Mayagüez. His research interests include architecture and protocol design for computer and communication networks, performance analysis, network security, and wireless communications.

Michel Kadoch (S’67–M’77–SM’04) received the B.Eng. degree from Sir George Williams University, Montreal, QC, Canada, in 1971, the M.Eng. degree from Carleton University, Ottawa, ON, Canada, in 1974, the M.B.A. degree from McGill University, Montreal, in 1983, and the Ph.D. degree from Concordia University, Montreal, in 1991. He is currently a Full Professor with the Ecole de Technologie Superieure, Universite du Quebec, Montreal. He is also an Adjunct Professor with Concordia University. He serves as a reviewer for a number of journals and conferences, as well as for the Natural Sciences and Engineering Research Council of Canada grants. His current research interests include cross-layer design, reliable multicast in wireless ad hoc, and WiMAX networks. He has publications and patents in the aforementioned areas.