On the Load Balancing of Multicast Routing in Multi-Channel Multi-Radio Wireless Mesh Networks with Multiple Gateways

Transcription

1 206 IJCSNS International Journal of Computer Science and Networ Security, VOL.17 No.3, March 2017 On the Load Balancin of Multicast Routin in Multi-Channel Multi-Radio Wireless Mesh Networs with Multiple Gateways Majid Asadi Shahmirzadi 1, Mehdi Dehhan 2, Abdorasoul Ghasemi 3 1 Department of Computer Enineerin, Science and Research Branch, Islamic Azad University, Tehran, Iran 2 Computer Enineerin Department, Amirabir University of Technoloy, Tehran, Iran 3 Faculty of Computer Enineerin, K. N. Toosi University of Technoloy, Tehran, Iran Summary Effective multicast routin schemes alleviate potential conestion on nodes and channels, thereby improvin networ throuhput. This paper studies the problem of findin the shortest load-balanced multicast tree with minimum channel utilization in response to a multicast communication request in multi-channel multi-radio wireless mesh networs with multiple ateways. We propose an optimization framewor based on mixed inteer prorammin, which minimizes the maximum of the channel utilization in multicast communication. The proposed Load-balanced Multicastin with Multiple Gateways (LMMG) framewor benefits from an appropriate node selection in the construction of the multicast tree. LMMG is efficient as it i) constructs the paths between the source and receivers throuh employment of multiple ateways, thus substantially reducin the interference and usae of resources; ii) simultaneously performs tree construction as well as channel and ateway selection processes. The performance of the proposed scheme is successfully compared to that of an existin method on different simulation scenarios. The results of our simulations demonstrate that balancin channel utilization dramatically enhances the networ performance. Key words: Wireless mesh networ, Multi-radio, multi-channel, Multi-ateway, Multicast routin, Load balancin 1. Introduction Wireless Mesh Networ (WMN) is an emerin and promisin networ technoloy that prepares a broadband wireless Internet access for end users. Due to its desirable characteristics such as self-confiuration, low installation and maintenance cost, ease of deployment as well as intrinsic path diversity of the multi-hop wireless bacbone, wireless mesh networs are considered as one of the best solutions for a wide rane of applications [1]. Major components of a wireless mesh networ are ateways, mesh routers, and mesh clients. The ateways exchane the data pacets between the wireless mesh bacbone and the wired Internet bacbone. The mesh routers route the data pacets throuh wireless lins to deliver them to Internet or the mesh clients as end users [2]. WMN as a broadband internet access suffers from the deradation of throuhput due to the underlyin interference. Employment of the technoloy of Multi-Channel Multi-Radio (MCMR) is an effective approach to alleviate this critical problem. In this paradim, the ateways and mesh routers are equipped with multiple radios adjusted to several available channels. The equipped routers simultaneously allow transmittin or receivin pacets throuh different channels. Even thouh the limited number of the available channels hinders complete removal of the interference, an efficient channel selection can sinificantly improve networ performance [1], [3]. Maximum networ performance cannot be attained unless load balancin is considered. In eneral, the load balancin problem can be viewed from two perspectives: node load balancin, and channel load balancin [4]. In the former, the conestion on a node maes a lon time waitin for data transmission or reception. This clearly diminishes throuhput of the entire networ. In the latter, which is the focus of this paper, we distribute the load between the channels as to enhance the networ performance. This enhance performance maes the networ more responsive to the future communication requests. On the other hand, the demand for roup communication to use multimedia services has substantially increased in the last recent years, main the multicast communication a critical research topic in WMN. A multicast routin directs data pacets from the source to multiple receivers. Due to the clear inapplicability of unicast-based performin in this scenario, an efficient multicast routin mechanism is essential. Despite the wide rane of research devoted to multicastin in WMN, the literature on MCMR is still limited. In addition, ateways are sinificantly important in WMN, as they deliver the networ traffic to the Internet. Since a sinle ateway desin maes the ateway a bottlenec, multiple ateways are needed to distribute the entire traffic load. Therefore, the ateway selection can directly improve load balancin in the networ [2]. In multicastin, in particular, multicast receivers attain the same data throuh different ateways. Hence, efficient ateways selection is more crucial in multicast communication to enhance networ performance [5]. A common way to fulfill a multicast communication is the construction of a multicast tree. This entails the Manuscript received March 5, Manuscript revised March 20, 2017.

2 IJCSNS International Journal of Computer Science and Networ Security, VOL.17 No.3, March specification of the constitutive nodes and the channels for the communication. Dependin on their communication load, channels may be unequally utilized, a fact that paves the way for maximum utilization of channels. However, in wireless environments, the channel utilization relative to a node is dependent on the location of the node and its interferin nodes [4][6]. In our scheme, we plan to distribute evenly channel utilization by considerin the interference and consumed capacity of the channel for each node in the networ. Our oal is to study the problem of findin the shortest load-balanced multicast tree with minimum channel utilization in response to a multicast communication request in multi-channel multi-radio wireless mesh networs with multiple ateways. To this end, we propose an optimization framewor, which we call the Load-balanced Multicastin with Multiple Gateways (LMMG). LMMG features a mixed inteer prorammin formulation and finds the shortest load-balanced multicast tree when the networ receives a multicast request. The efficiency and performance of the proposed scheme are successfully compared to an existin method. The rest of the paper is oranized as follows. Section 2 briefly reviews the related wors. Section 3 elaborates on the networ model and the problem statement. Section 4 presents the optimization model, and Section 5 demonstrates the performance of LMMG under two different scenarios. Finally, summary and conclusions are iven. 2. Related Wor The subjects of multicastin and load balancin in wireless mesh networs (WMN) have been studied mostly separately in the literature. In particular, rather less attention has been iven to employin multi-channel multi-radio in the networ model. In the followin, we briefly review multicast tree construction and load balancin in WMN. Galvez et al. [7] proposed an adaptive online load-balancin protocol for multi-ateway WMNs. Their protocol balances the load between ateways based on the current networ conditions. It identifies the conested domains and reroutes flows to other domains. The load of a ateway is measured as the number of flows served by the ateways. If ateways serve a similar number of flows, then the networ capacity is shared which provides an improved performance. Zhao et al. [8] proposed a Gateway-cluster based Load Balancin Multicast alorithm (GLBM) to achieve Quality of Service (QoS) by avoidin uneven traffic load. They used the IFQ (Networ Interface Queue) lenth to examine a certain node s traffic load. In their scheme, the path with the lowest IFQ areation is specified for the multicast tree in order to prevent reusin the heavy load nodes. Zhao et al. [9] later considered center node load balancin. Their main idea is based on the fact that a node with a heavy traffic maes the neihborin nodes busy too. Therefore, they consider the busy neihborhood rather than the busy node alone in bypassin the traffic. Pourfahar and Rahmani [10] presented a neural networ model to predict route disjoint or node fault to control conestion. Authors also presented a QoS multicast routin framewor for WMNs to solve the problem of load balancin and enhance the QoS in multicast communication amon ateways and receivers. Zen et al. [11] proposed two multicast alorithms, namely, the Level Channel Assinment (LCA) and Multi-Channel Multicast (MCM) schemes, for the optimization the throuhput in MRMC WMNs. LCA alorithm first constructs a multicast tree based on breadth-first search to minimize the hop count distance from the source to the receivers. Then, it uses a dedicated stratey to assin channels to the tree to reduce the interference. In addition, MCM scheme builds the tree to minimize the number of the relayin nodes and the hop count distances between the source and the destinations. It also reduces the interference by exploitin all the available channels. Liu and Liao [12] considered the joint problem of channel assinment and multicast routin. The oal is to find the minimum cost multicast tree in MRMC WMNs. Here, the cost of a multicast tree is defined as the number of transmissions required to deliver data from source to receivers. An optimal model is presented for solvin this problem. They proposed a near optimal alorithm to construct the minimum transmission trees considerin their interference. Asadi Shahmirzadi et al. [13] proposed a cross-layer optimization framewor for joint channel assinment and multicast tree construction in MCMR WMNs. The cost function of the problem is the total number of the forwardin nodes and the interference. The proposed model incorporates multiple ateways and partially overlappin channels into the scheme, and offers optimal solution. Chiu and Yeun [6] proposed a call admission control for multicast in MCMR WMN. To increase the call acceptance rate, load balancin is fulfilled by minimizin the carried load on the overloaded channels and nodes. They proposed an optimal model as well as a heuristic approach to solve the problem. Avoh and Mirjalily [4] suested a heuristic alorithm to construct a load-balanced multicast tree in MRMC WMN. In this alorithm, the traffic load on mesh nodes is formulated as a conestion factor related to lins. In addition, utilizin wireless broadcast advantae was also devised in the lin selection alorithm. These two considerations define a cost function in which the cost of each lin is computed to enerate minimum cost tree.

3 208 IJCSNS International Journal of Computer Science and Networ Security, VOL.17 No.3, March 2017 As elaborated above, the subjects of multicastin and load balancin in wireless mesh networs have been viewed mostly separately in the literature. Also, major attention has been iven to employin sinle-channel sinle-radio in the networ model. In contrast, we consider the problem of findin the shortest load-balanced multicast tree with minimum channel utilization in response to a multicast communication request in multi-channel multi-radio wireless mesh networs with multiple ateways. 3. Networ Model and Problem Statement In the followin, the networ model and problem statement are elaborated. 3.1 Networ Model We consider the static, multi-channel, multi-radio, multi-ateway, wireless mesh networ shown in Fi. 1. Let C denote the set of channels, G the set of ateways, and R the set of receivers in the networ, with the assumption that receivers not only can receive data, but also can forward the data. Let the communication raph Gr = (V, E) represent this networ, where V and E are the set of mesh routers and wireless lins, respectively. Mesh routers are equipped with multiple radios, each of which can be tuned to a distinct non-overlappin channel. In addition, wireless lins are created when two nodes are located within the transmission rane of each other and have a radio tuned to the same channel. We employ a receiver conflict avoidance interference model [5] in this paper. Under this model, a data transmission from one node to another is successful if there is no other node transmittin data on the same channel within the interference rane of the taret node. Here, we assume that i) the channel assinment is iven in the networ, ii) all radios have a common transmission rane, T r, and a common interference rane I r ( T r ), and iii) each mesh router has radios tuned to distinct channels, as to avoid self-interference [14]. 3.2 Problem Statement The oal is to find the shortest load-balanced multicast tree with minimum channel utilization in response to a multicast communication request. Our approach to this problem is to convert it into a min-max problem. Thus, we minimize the lonest path in the tree and the maximum Fi.1 A schematic of a Multi-Channel Multi-Radio Multi-Gateway Wireless Mesh Networ (MCMRMG WMN). The number on each lin represents the lin s channel number. channel utilization in the networ. The lonest path in the tree is defined as h max = max i R, G {h v }, (1) where the inteer h v ε[0, N n ] quantifies the distance between the node v from the ateway. Here, N n is the total number of nodes. Also, the maximum channel utilization is defined by U max = max v V, C {U v }, (2) where U v denotes the usae of the channel relative to node v. The parameter U v can be written as U v = E v + T v, ( v V, C), (3) where E v is the occupied bandwidth of channel relative to node v for the existin lins in the networ, and T v is that for the tree construction, respectively. The parameter E v is defined as E v = i,j, (i I v, j V, C), (4) A ij L ij where set I v contains the node v and those in the interference rane of node v. The binary variable A ij, when equals one, determines if channel is assined to the lin (i, j). In addition, the parameter L ij denotes the load capacity, which the channel shares with the lin (i, j). In addition, the parameter T v is defined as

4 IJCSNS International Journal of Computer Science and Networ Security, VOL.17 No.3, March T v = i,j A ij B m, (i I v, j V G, C), (5) where l ij is a binary variable, which, if equals one, indicates that the lin (i, j) on multicast tree is connected to the ateway. Here, B m is the bandwidth required by the multicast communication. 4. LMMG: An Optimization Framewor We cast the problem elaborated in section 3 into an optimization framewor. In the followin, we propose a solution framewor, which we call the Load-balanced Multicastin with Multiple Gateways (LMMG). The objective function and the constraints of this framewor are elaborated as follows. 4.1 Objective Function LMMG minimizes the weihted sum of the maximum channel utilization (iven by eq. (1)) and the lonest path in the multicast tree (iven by eq. (2)). That is, = h max /N n + U max /C a, (6) where N n is the number of nodes in the networ, and C a is the channel capacity. Note that the parameters C a and N n are chosen to effectively combine the contributions of channel utilization and depth of the multicast tree in the definition of the objective function. 4.2 Constraints It is clear that no channel can be loaded more than its capacity, C a. Thus, the first constraint in the proposed optimization framewor is U i C a, (i V, C). (7) In addition, the objective function iven by (6) assumes that the parameters U max and h max attain maximum values, the fact that constitutes the second set of constraints. That is, U max U i, (i V, C), (8) h max h i, (i R, G). (9) Further constraints arise due to the construction of the multicast tree. This is common to the problem studied by Asadi Shahmirzadi et al. [13]. Thus, we tae advantae of the exitin derivations. Asadi Shahmirzadi et al. [13] presented constraints arisin from the usae of ateways, receivers, relayin nodes, and consideration of loop prevention. The constraints for ateways are i, =0, (i V, p, G), (10) l ip l i i, 1, (i V G, G), (11) l i i, r, (i V G), (12) l i, i, = i,,p l pi (i V G, p, G). (13) The constraints for receivers are i, = 1, (i V, j R, G), (14) i, 0, (i R, j V, G). (15) The constraints for relayin nodes are i, 1, (i V, j V G R, G), (16) i l jx x, (i V, j V G R, x V G, G), (17) ( 1 + i Z ij ) i l x jx, (i V, j V G R, x V G, G), (18) Z ij, (i, j V, G). (19) where Z ij = { 1, D ij < T r, (i, j V). (20) 0, else In eq. (20), D ij is the Euclidean distance between the nodes i and j. Finally, the loop prevention constraints are h = 1, ( G), (21) i h j, (i V, j V G, G), (22) N n i h j, (i V, j V G, G), (23) h j h i N n + (N n + 1), (i, j V, G), (24) h j h i N n (N n 1), (i, j V, G). (25) 5. Simulation Results The proposed model is implemented in Advanced Interated Multidimensional Modelin Software (AIMMS version 3.0.1) [15]. We use the built-in CPLEX 12.2 solver and the branch and cut method for mixed inteer prorammin. We conduct two sets of simulations to evaluate the performance of the proposed scheme. In particular, we consider two sets of rid topoloy sizes in our MCMRMG WMN simulations. One has 16 nodes arraned within a square of 400m 400m, and the other has 25 nodes arraned within a square of 500m 500m. Additionally, the number of radios per node is randomly determined and is limited to 3. Finally, three non-overlappin channels are employed in the networ, where the transmission and interference ranes of the radios are set to 100m and 200m. In the followin simulations, we employ two metrics, namely, the standard deviation of the channels utilization, and the maximum heiht of the multicast tree. The less the value of the former, the more the balance of the channel utilization. On the other hand, the smaller the latter, the less the delay in the multicast communication. Scenario 1: In the first scenario, we study the effect of multicast roup size on the alorithm s performance. We

5 Channel Utilization Standard Deviation (CUSD) Channel Utilization Standard Deviation (CUSD) Heiht of multicast tree (hop count) 210 IJCSNS International Journal of Computer Science and Networ Security, VOL.17 No.3, March 2017 consider a networ with 16 mesh routers and a varyin number of receivers ranin from 2 to 8. In each multicast roup, the simulations have been repeated for different permutations of nodes. Multicast tree is built with up to two ateways usin two approaches: i) LMMG, and ii) the Shortest Path Tree (SPT), which connects the source to the receivers in the shortest path. The simulations are started with an initially loaded system, with each lin loaded randomly up to 10% of the capacity of its channel. In addition, the bandwidth required for the multicast request is set to 5% of the channel capacity. As illustrated in Fi. 2, LMMG provides a better performance compared to SPT. Recall that the less the channel utilization standard deviation, the better the performance. In particular, improvements are more pronounced when multiple ateways are employed in the multicast tree construction. In addition, Fi.3 presents the effect of the size of the multicast roup on the heiht of the multicast tree for the two schemes LMMG and SPT. As expected, SPT scheme constructs an optimal multicast tree with minimum heiht. LMMG not only does obtain the same result, but also returns a load-balanced multicast tree. Also, results follow intuition in that the more the ateways used in the networ, the shorter the multicast tree, thus the faster the multicast communication. Scenario 2: In this scenario, we expand the size of the networ to have 25 mesh routers, and the size of the multicast roup from 3 to 12 nodes. We confiure the networ with the same parameters similar to the previous scenario. As illustrated in Fis. 4 and 5, simulations reveal that LMMG retains its enhanced performance compared to SPT Fi. 3 Comparison of LMMG and SPT schemes on the heiht of the multicast tree for a 16-node networ Fi.4 Comparison of LMMG and SPT schemes on the Channel Utilization Standard Deviation for a 25-node networ Fi. 2 Comparison of LMMG and SPT schemes on the Channel Utilization Standard Deviation for a 25-node networ. Note that the number 1 in stands for the number of ateways. This naturally extends to other cases.

6 Heiht of multicast tree (hop count) IJCSNS International Journal of Computer Science and Networ Security, VOL.17 No.3, March Fi.5 Comparison of LMMG and SPT schemes on the heiht of the multicast tree for a 25-node networ 6. Conclusion We studied the problem of findin the shortest load-balanced multicast tree with minimum channel utilization in response to a multicast communication request in multi-channel multi-radio wireless mesh networs with multiple ateways. We proposed an optimization framewor based on mixed inteer prorammin, which minimizes the maximum of the channel utilization in multicast communication. Our framewor benefits from an appropriate node selection in the construction of the multicast tree. The framewor is efficient as it i) constructs the paths between the source and receivers throuh employment of multiple ateways, thus substantially reducin the interference and usae of resources; ii) simultaneously performs tree construction as well as channel and ateway selection processes. We demonstrated the superior performance of the proposed scheme compared to an existin method on different simulation scenarios. The results of our simulations demonstrated that balancin channel utilization dramatically enhances the networ performance. Also, the results confirm that the performance of LMMG is retained as the networ size is increased. References [1] I.F.Ayildiz, X.Wan, and W. Wan, Wireless mesh networs: a survey, Computer networs, Vol. 47, No. 4, pp , [2] V.C. Bores, M. Curado, and E. Monteiro, Cross-layer routin metrics for mesh networs: Current status and research directions, Computer Communications, Vol. 47, No. 6, pp , [3] M. Alicherry, R. Bhatia, and L. E. Li,. Joint channel assinment and routin for throuhput optimization in multi-radio wireless mesh networs, In Proceedins of the 11th annual international conference on Mobile computin and networin, pp ACM, [4] A. Avoh, and G. Mirjalily, Load-balanced multicast tree routin in multi channel multi radio wireless mesh networs usin a new cost function, Wireless personal communication, Vol. 69, No. 1, PP , [5] O. Bahban Karimi, J. Liu, and Z. Li, Multicast with cooperative ateways in multi-channel wireless mesh networs, Ad Hoc Networs, Vol. 13, PP , [6] H. S. Chiu, and K. L. Yeun, Maximizin multicast call acceptance rate in multi-channel multi-interface wireless mesh networs, IEEE Transactions on Wireless Communications, Vol. 9, No. 8, PP , [7] J. J. Galvez, P. M. Ruiz, and A. F. Sarmeta, Responsive on-line ateway load-balancin for wireless mesh networs," Ad Hoc Networs, Vol. 10, No. 1, PP , [8] L. Zhao, A. Y. Al-Dubai, and G. Min, GLBM: A new QoS aware multicast scheme for wireless mesh networs, Journal of Systems and Software, Vol. 83, No. 8, PP , [9] L. Zhao, A. Y. Al-Dubai, and G. Min, An efficient neihbourhood load routin metric for wireless mesh networs, Simulation Modellin Practice and Theory, Vol. 19, No. 6, PP , [10] E. Pourfahar, and A. M. Rahmani, A hybrid QoS multicast framewor-based protocol for wireless mesh networs. Computer Communications, Vol. 33, No.17,PP , [11] G. Zen, B. Wan, Y. Din, L. Xiao, and M. W. Muta, Efficient multicast alorithms for multichannel wireless mesh networs, IEEE Transactions on Parallel and Distributed Systems, Vol. 21, No. 1, PP , [12] T. Liu, and W. Liao, Multicast routin in multi-radio multi-channel wireless mesh networs, IEEE Transactions on Wireless Communications, Vol. 9, No. 10, PP , [13] M. Asadi Shahmirzadi, M. Dehhan, and A. Ghasemi, An optimization framewor for multicastin in MCMR wireless mesh networ with partially overlappin channels, Wireless Networs,PP. 1-19, 2016.[Online]. Available: /s x [14] V. Sarasvathi, N. C. Iyenar, and S. Snehanshu, An efficient interference aware partially overlappin channel assinment and routin in wireless mesh networs, International Journal of Communication Networs and Information Security, Vol 6, No. 1, pp , [15]