Your SlideShare is downloading. ×
Bandwidth efficient video multicasting in multiradio multicellular wireless networks
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Introducing the official SlideShare app

Stunning, full-screen experience for iPhone and Android

Text the download link to your phone

Standard text messaging rates apply

Bandwidth efficient video multicasting in multiradio multicellular wireless networks

446
views

Published on

Published in: Technology, Business

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
446
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
8
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008 275 Bandwidth Efficient Video Multicasting in Multiradio Multicellular Wireless Networks De-Nian Yang, Member, IEEE, and Ming-Syan Chen, Fellow, IEEE Abstract—In this paper, we propose a new mechanism to select the cells and the wireless technologies for layer-encoded video multicasting in the heterogeneous wireless networks. Different from the previous mechanisms, each mobile host in our mechanism can select a different cell with a different wireless technology to subscribe each layer of a video stream, and each cell can deliver only a subset of layers of the video stream to reduce the bandwidth consumption. We formulate the Cell and Technology Selection Problem (CTSP) to multicast each layer of a video stream as an optimization problem. We use Integer Linear Programming to model the problem and show that the problem is NP-hard. To solve the problem, we propose a distributed algorithm based on Lagrangean relaxation and a protocol based on the proposed algorithm. Our mechanism requires no change of the current video multicasting mechanisms and the current wireless network infrastructures. Our algorithm is adaptive not only to the change of the subscribers at each layer, but also the change of the locations of each mobile host. Index Terms—Multicast, layer-encoded video, heterogeneous wireless networks. Ç1 INTRODUCTIONT HE success of wireless and mobile communications in the 21st century has resulted in a large variety ofwireless technologies such as second- and third-generation quality. The layers are distributed to receivers via multicast channels in wireless networks [6]. Previous works on layer-encoded video multicast mainlycellulars, satellite, Wi-Fi, and Bluetooth. The heterogeneous focus on deciding the set of subscribed layers for thewireless networks combine various wireless networks and receivers with heterogeneous requirements and networkprovide universal wireless access. Users in the heteroge- conditions. McCanne et al. [7] propose a protocol to deliverneous wireless networks are usually covered by more than the required layers to each of the receivers. Vickers et al. [8]one cell to avoid connection drop and service disruption. In adaptively adjust the number of layers and the bit rate ofaddition, more mobile terminals in the wireless networks each layer according to the network condition. Kim andare likely to own multiple wireless technologies. Therefore, Ammar [9] develop a quality adaptation scheme thatthe heterogeneous wireless networks provide the mobile maximizes the perceptual video quality through minimiz-hosts with many choices for the cell and the wireless ing quality variation. Liu et al. [10] find the optimal numbertechnologies to access the Internet. of layers and the amount of bandwidth for each layer to Video delivery in wireless networks is becoming an maximize the total utilities of all mobile hosts. Zhao et al.important multimedia application due to the proliferation [11] determine the optimal power allocation and theof the Web-based services and the rapid growth of wireless transmission rate of the base station to deliver each layercommunication devices. For a video stream delivered to a of a video stream in CDMA networks. Therefore, previoussingle mobile host, a video server can adaptively adjust the works assume that the network operators have selected theencoder to accommodate the delay, jitter, and packet loss of set of cells to multicast a video stream. Moreover, each cellthe networks [1], [2], [3]. For a video stream delivered to must multicast the layers of a video stream in themultiple mobile hosts with diverse requirements and progressive manner. In other words, a cell cannot delivernetwork conditions, the video stream can be encoded at the enhancement layers without multicasting the base layerthe highest resolution and divided into layers such that each of a video stream.receiver can decode the stream on the preferred rate and In this paper, different from the previous works, weresolution with a set of layers [4], [5]. The most significant focus on the selection of the cells and the wirelesslayer, that is, the base layer, contains the data representing technologies to multicast each layer of a video stream inthe most important features of the video, whereas the the heterogeneous wireless networks, where each selectedadditional layers, that is, the enhancement layers, contain cell can deliver only a subset of layers of the video stream todata that progressively refine the reconstructed video reduce the total bandwidth consumption. In our approach, each mobile host can select a different cell with a different. The authors are with the Department of Electrical Engineering, National wireless technology to subscribe each layer of a video Taiwan University, #1 Roosevelt Rd., Sec. 4, Taipei, Taiwan 106, ROC. stream. Consider Fig. 1 as an example, where A, B, C, and D E-mail: {dnyang, mschen}@cc.ee.ntu.edu.tw. are the mobile hosts. Here, we assume that the bandwidthManuscript received 11 Sept. 2005; revised 29 July 2006; accepted 17 Dec. allocated to each Universal Mobile Telecommunications2006; published online 28 June 2007. System (UMTS) and each Wi-Fi cell can support one andFor information on obtaining reprints of this article, please send e-mail to:tmc@computer.org, and reference IEEECS Log Number TMC-0272-0905. two layers of a video stream. The solid line connecting aDigital Object Identifier no. 10.1109/TMC.2007.70725. mobile host and a base station represents that the mobile 1536-1233/08/$25.00 ß 2008 IEEE Published by the IEEE CS, CASS, ComSoc, IES, & SPS Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 2. 276 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008Fig. 1. Comparison of two different selections of cells and technologies to deliver each layer of a video stream. (a) Each mobile host subscribes alllayers from a single cell. (b) Each mobile host can subscribe each layer from a different cell.host subscribes Layer 1, the base layer, of a video stream manner, and the important merit of algorithm LAGRANGEfrom the cell. The long dash line and the short dash line enables us to design a network protocol accordingly.correspond to Layer 2 and Layer 3, the enhancement layers, Second, the algorithm adapts to the change of theof the video stream. For the network operators, each layer subscribers at each layer of a video stream and the changeconsumes less network bandwidth in our approach. The of locations of each mobile host. The algorithm iterativelythree cells multicast Layer 1 in Fig. 1a. In Fig. 1b, however, reduces the total bandwidth consumption according to thethe two Wi-Fi cells do not multicast Layer 1 because mobile current subscribers of each layer at each location. Third, thehosts A and B subscribe the layer from the UMTS cell. Note algorithm provides the lower bound on the total band-that mobile hosts A and B subscribe Layer 2 and 3 from a width cost of the optimal selected cells and the wirelesssingle Wi-Fi cell to minimize the total bandwidth consump- technologies. For a video stream with a large number oftion in Fig. 1b. For the users, each mobile host can subscribe subscribers, the lower bound obtained by our algorithmmore layers in our approach. Mobile hosts A and B provides a benchmark to compare with any algorithm forsubscribe two layers if they subscribe all layers from only the problem since using the ILP formulation here to findthe Wi-Fi cells in Fig. 1a. On the contrary, mobile hosts A the optimal solution is computationally infeasible. Inand B subscribe three layers in Fig. 1b. They subscribe addition, the algorithm requires no change of the currentLayer 1 from the UMTS cell and Layers 2 and 3 from the Wi- video multicasting mechanisms and the current wirelessFi cells. Subscribing the layers from the cells with multiple network infrastructures.wireless technologies may consume more power. Some Algorithm LAGRANGE can be regarded as a reroutingmobile hosts are concerned with the video quality but mechanism. The network may use different cells to multi-others are concerned with the power consumption. To cast each layer of a video stream when the networkcapture the heterogeneity of users, therefore, the number of condition changes. Note that rerouting mechanisms havewireless technologies allowed to be used by each mobile been designed for unicast communications in circuithost can be different in this paper. switched networks [18], optical networks [19], and satellite Explicitly, we formulate in this paper the selection of the networks [20] to reduce the total bandwidth consumption.cell and the wireless technology to multicast each layer of a However, designing a rerouting mechanism for layer-video stream as an optimization problem, which is referred encoded video multicast in the heterogeneous wirelessas the Cell and Technology Selection Problem (CTSP) in the networks is more difficult. First, the selection of cells for anyheterogeneous wireless networks for layer-encoded video two layers is correlated because each mobile host can usemulticasting. The objective of the problem is to minimize only a limited number of cells and wireless technologies.the total bandwidth cost of the selected cells and the Second, we show that the multicast communication makeswireless technologies. We design a mechanism, which our problem become NP-hard in Section 2. Since ourincludes an Integer Linear Programming (ILP) formulation, mechanism can be regarded as a rerouting mechanism, aa distributed algorithm, and a network protocol to solve mobile host in our mechanism needs the sophisticatedthe CTSP. The network operators can use our ILP handover mechanisms such as those in the literature [21],formulation for network planning. We show that the [22], [23], [24] to subscribe the video stream from the properproblem is NP-hard and design an algorithm LAGRANGE, cells in different situations.which is based on Lagrangean relaxation [12] on our ILP Overall, the contributions of this paper are many-fold:formulation. Lagrangean relaxation has been widely usedin various network design, control, and pricing problems . We consider the heterogeneity of mobile hosts. Some[13], [14], [15], [16], [17], and the convergence of Lagran- mobile hosts would like to improve the videogean relaxation has been proved in the literature [12]. We quality by subscribing more layers from more cellsadopt Lagrangean relaxation in our algorithm instead of with proper wireless technologies. Other mobileother optimization techniques due to the following reasons: hosts may concern the power consumption andFirst, our algorithm decomposes the original problem into thereby subscribe a limited number of layers with amultiple subproblems such that each subproblem can be single cell.solved by each mobile host individually. In other words, . For each wireless technology, our mechanismthe algorithm can be implemented in the distributed reduces the number of cells to multicast a video Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 3. YANG AND CHEN: BANDWIDTH EFFICIENT VIDEO MULTICASTING IN MULTIRADIO MULTICELLULAR WIRELESS NETWORKS 277 stream by clustering the mobile hosts. Therefore, we available wireless technology. Moreover, to consider the can reduce the wireless bandwidth consumption heterogeneity of mobile hosts, the number of wireless even when the operators own only one wireless technologies used by each mobile host can be different. In network. this paper, we concern only the wireless bandwidth . Our mechanism chooses the proper wireless tech- consumption of each multicast group, and we assume that nologies to minimize the total bandwidth cost of all the networks have an abundant wireline bandwidth to wireless networks. For a set of nearby mobile hosts, connect to each cell. In addition, we assume the base station our algorithm minimizes the bandwidth cost by of a cell needs to multicast a layer of a video stream if using a single large cell or multiple smaller cells to at least one mobile host in the cell subscribes the layer from serve these mobile hosts according to the number of the cell. In this paper, a cell covers a mobile host if the mobile hosts, the locations of mobile hosts, and the mobile host is within the transmission range of the base bandwidth cost of each wireless technology speci- station of the cell, and the cell in this case is a covering cell of fied by the network operators. the mobile host. A cell is a candidate cell if it covers at least . Our mechanism is flexible since the bandwidth cost one mobile host. A wireless technology is a candidate of each cell can be assigned with no restriction, and technology for a mobile host if the mobile host can access the cost of each cell can be adjusted dynamically. the Internet with the wireless technology. For simplicity, the The flexible cost model enables the network selection of the cell for each mobile host means the selection operators to balance the load of wireless cells with of both the cell and the wireless technology in the rest of the same or different technologies by adjusting the this paper. cost [25]. The notation in this paper is summarized as follows: . Our mechanism is transparent to the video multi- casting mechanisms. Each mobile host subscribes the . C: This is the set of cells in the heterogeneous video stream with the current video multicasting wireless network. mechanisms after the mobile host selects the cell and . L: This is the set of layers of a layer-encoded video the wireless technology according to our mechan- stream. ism. We thereby require no modification on the . M: This is the set of mobile hosts in the network. current video multicasting mechanisms. . T : This is the set of wireless technologies. . Our mechanism requires no modification on the . Cm : This is the set of cells that cover mobile host m current wireless network infrastructures. The algo- and have good channel condition, m 2 M, Cm  C; if rithm is implemented in only the mobile hosts, and mobile host m selects cell c manually, we let Cm the mobile hosts cooperatively select the cells and contain only a cell c. the wireless technologies. . Ct : This is the set of cells with wireless technology t, . Our mechanism is adaptive. Our algorithm adapts to t 2 T , Ct  C. the change of the subscribers at each layer of a video . Lm : This is the set of layers subscribed by mobile stream, the change of locations of each mobile host, host m, m 2 M, Lm  L. and the change of the bandwidth cost of each cell. In . Mc : This is the set of mobile hosts covered by cell c, addition, our mechanism enables each mobile host to c 2 C, Mc  M. either automatically or manually choose the cell and . Tm : This is the set of wireless technologies that can the wireless technology. For example, when a mobile be used by mobile host m, m 2 M, Tm  T . host manually chooses a cell, other nearby mobile . bc;l : This is the bandwidth cost of cell c to deliver hosts adaptively attach to the cell to reduce the total layer l of the video stream, c 2 C, l 2 L. bandwidth consumption. . tm : This is the number of wireless technologies that The rest of the paper is summarized as follows: We can be used by mobile host m, m 2 M.describe our assumption, present our ILP formulation, and . m;c;l : This is the Lagrange multiplier for mobileshow that the CTSP is NP-hard in Section 2. We propose host m with respect to cell c and layer l, m 2 M,algorithm LAGRANGE based on Lagrangean relaxation c 2 Cm , l 2 Lm .and the corresponding protocol in Section 3. We show our We use ILP to model the CTSP. The ILP formulation canexperimental results in Section 4. Finally, we conclude this find the optimal cells and technologies to multicast eachpaper in Section 5. layer of a video stream. The network operators can use our ILP formulation for network planning in the heterogeneous wireless networks. Our ILP formulation has the following2 PROBLEM DESCRIPTION variables:In this paper, we consider the CTSP in the heterogeneous . m;c;l : This is a binary variable; m;c;l is one if mobilewireless networks for layer-encoded video multicasting. host m selects cell c for layer l of the video stream,Video streams are delivered with one hop from the base m 2 M, c 2 Cm , l 2 Lm .station to each mobile host. The problem is to select the cells . m;t : This is a binary variable; m;t is one if mobileand the wireless technologies to multicast each layer of a host m selects any cell with technology t, t 2 Tm .video stream such that the total bandwidth cost is . m;c : This is a binary variable; m;c is one if mobileminimized. The problem is subject to the constraint that host m selects cell c for any subscribed layer, m 2 M,each mobile host can select at most one cell for each c 2 Cm . Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 4. 278 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008 .c;l : This is a binary variable; c;l is one if cell c Therefore, the Minimum Set Cover problem is the same as delivers layer l of the video stream; namely, at least the CTSP if each mobile host subscribes only a single layer one mobile host subscribes layer l from cell c, c 2 C, of the video stream and owns only one candidate wireless l 2 L. technology. In other words, the CTSP is more difficult than The objective function of our ILP formulation is given as the Minimum Set Cover problem because our problemfollows: allows each mobile host to subscribe multiple layers and XX use multiple wireless technologies. min bc;l  c;l : c2C l2L 3 DESIGN OF ALGORITHM LAGRANGEAlthough we do not maximize the number of subscribedlayers for each mobile host in our problem, the objective In this section, we propose algorithm LAGRANGE, whichfunction here minimizes the total wireless bandwidth is based on Lagrangean relaxation on our ILP formulationconsumption to deliver the current subscribed layers of proposed in Section 2. The algorithm can be implemented inthe stream to the mobile hosts. Therefore, with the objective the distributed manner on only mobile hosts. The algorithmfunction, each mobile host is able to subscribe more layers adapts to the change of the subscribers of each layer, thewith a fixed amount of wireless bandwidth and the change of the location of each mobile host, and the changenetworks thereby can support more video streams. Our of the bandwidth cost of each cell. In addition, theILP formulation includes the following constraints: algorithm provides a lower bound on the total bandwidth X cost of the optimal solution to the CTSP. m;c;l ¼ 1; 8m 2 M; 8l 2 Lm ; ð1Þ The algorithm relaxes a constraint of our ILP formulation c2Cm and transfers the CTSP into the Lagrangean Relaxation m;c;l m;c ; 8m 2 M; 8c 2 Cm ; 8l 2 Lm ; ð2Þ Problem (LRP). The LRP owns a new objective function X with the Lagrange multipliers and fewer constraints such m;c 1; 8m 2 M; 8t 2 Tm ; ð3Þ c2Ct Cm that we can decompose the LRP into multiple subproblems, where each subproblem can be solved in a distributed m;c;l m;t ; 8m 2 M; 8t 2 Tm ; 8c 2 Ct Cm ; 8l 2 Lm ; ð4Þ X manner by the mobile hosts. The mobile hosts in our m;t tm ; 8m 2 M; ð5Þ algorithm cooperatively select the cells and the wireless t2Tm technologies for each layer of a video stream according to m;c;l c;l ; 8m 2 M; 8c 2 Cm ; 8l 2 Lm : ð6Þ the solutions to the subproblems. The solutions to the subproblems depend on the Lagrange multipliers of the Constraint (1) guarantees that each mobile host selects LRP, and we iteratively adjust the multipliers to change theone cell for each subscribed layer. Constraints (2) and (3) selected cells to iteratively reduce the total bandwidthenforce that each mobile host selects at most one cell for consumption. In the rest of the section, we solve the CTSPeach available wireless technology. Constraints (4) and (5) with the following steps:state that the number of technologies used by a mobile hostmust be bounded by tm . Each mobile host m can be . Transfer the CTSP into the LRP.associated with a different tm to consider the heterogeneity . Decompose the LRP into multiple subproblems andof mobile hosts. Constraint (6) guarantees that a cell needs solve each subproblem, respectively.to multicast a layer of the video stream if any covered . Select the cell and the wireless technology for eachmobile host subscribes the layer from the cell. In addition to subscribed layer of each mobile host according to thethe above constraints, the problem is also subject to the solutions to the subproblems.integrality constraints to enforce that m;c;l , m;t , m;c , and c;l . Reduce the total bandwidth cost of the selected cellsare all binary variables. We regard a set of selected cells that by iteratively updating the cost, the Lagrangeobey the above constraints as a feasible solution to the CTSP. multiplier, of each cell for each subscribed layer of Previous works for layered video multicasting [7], [8], each mobile host.[9], [10], [11] assume that all layers of a stream are deliveredin a single cell. In this paper, however, we show that 3.1 Transferring the CTSP and Solving the LRPutilizing different cells and technologies to deliver each Our algorithm relaxes constraint (6) in our ILP formulationlayer of a video stream can reduce the bandwidth to transfer the CTSP into the LRP, and the objective functionconsumption. In addition, constraints (4) and (5) are of the LRP is shown as follows:formulated for the applications that use multiple multicast XX X X Xgroups simultaneously, and we believe that video multi- min bc;l  c;l þ m;c;l ðm;c;l À c;l Þcasting nowadays is the only application that is bandwidth c2C l2L m2M c2Cm l2Lm " ! #demanding and conforms to the above requirement. XX X We show that the CTSP in the heterogeneous wireless ¼ min bc;l À m;c;l c;l c2C l2L m:c2Cm ;l2Lmnetworks for multicast communications is NP-hard because " #the Minimum Set Cover problem [26] is a special case of the X X X þ m;c;l m;c;l ;CTSP. In the Minimum Set Cover problem, each set is m2M c2Cm l2Lmassigned a cost and covers some elements. The problem isto select the sets with the minimum total cost such that where m;c;l is the Lagrange multiplier, m;c;l ! 0, 8m 2 M,every element is covered by at least one selected set. 8c 2 Cm , 8l 2 Lm . The Lagrange multiplier m;c;l is the cost Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 5. YANG AND CHEN: BANDWIDTH EFFICIENT VIDEO MULTICASTING IN MULTIRADIO MULTICELLULAR WIRELESS NETWORKS 279associated with cell c for layer l of mobile host m. The LRP Proof. We prove the theorem with reduction from Maximumincludes the constraints (1)-(5) of our ILP formulation. k-Facility Location (MFL) problem, which is NP-hard [26].Compared with the objective function of the CTSP, the The problem is given a set of vertices V in a completeobjective function of the LRP owns a new term correspond- graph, where each edge et;l connecting vertices t and l ising to the relaxed constraint (6). Intuitively, for any feasible associated with a profit pt;l . The  problem is to find a solution to the LRP that contradicts the relaxed constraint, subset of vertices V  V with V  k to maximize thenamely, m;c;l > c;l , the objective function punishes the following total profit:solution with a larger objective value. Besides, any feasible Xsolution to the CTSP must also act as a feasible solution to max pt;l : l2V t2Vthe LRP since the set of constraints of the LRP is a subset ofthe constraints of the CTSP. Therefore, when we adopt the For each instance in the MFL problem, we create theoptimal solution to the CTSP as the feasible solution both to corresponding instance of our second subproblem asthe LRP and the CTSP, the objective value of the LRP must follows: The instance in our second subproblem has onlybe no more than the objective value of the CTSP because the one mobile host m. The instance contains a set of layers Vnew term in the objective function of the LRP is nonpositive. of a video stream, and the set of technologies that can beTherefore, the objective value of the optimal solution to the accessed by m is also V . Each technology has only oneLRP must be no more than the objective value of the cell that can be selected by m. For each layer l and theoptimal solution to the CTSP. In other words, the optimal cell c with technology t, we assign Lagrange multipliersolution to the LRP provides a lower bound on the objective m;c;l as follows:value of the optimal solution to the CTSP, which is the totalbandwidth cost of the selected cells and the wireless m;c;l ¼ q À pt;l ;technologies in the heterogeneous wireless networks. where q is maxu;v2V pu;v . Each host m can use only We solve the LRP by decomposing the LRP into two k technologies, namely, tm ¼ k.subproblems. We divide the objective function and the We prove the theorem by showing that the total profitconstraints of the LRP into two parts, where each subpro- of any MFL instance is p if and only if jV j  q À p is theblem owns one part of the objective function and the total cost in the corresponding instance of our secondconstraints. The variables in the two subproblems are subproblem. We first prove the sufficient condition. Ifmutually independent such that we can solve each sub- the set of vertices selected in the MFL instance is V , thenproblem individually, and the solution to the LRP is just we let mobile host m select the corresponding set ofthe combination of the solutions to the two subproblems. technologies V . Therefore, for each layer l, the cell The objective function of the first subproblem is the first selected by mobile host m for layer l must have thepart of the objective function in the LRP: following cost: ! XX X min m;c;l ¼ min q À pt;l ¼ q À max pt;l : min bc;l À m;c;l c;l : t:t2V ;c2Ct t2V t2V c2C l2L m:c2Cm ;l2Lm Therefore, the total cost of the instance in our second The first subproblem has only an integrality constraint subproblem is given as follows:enforcing that c;l must be a binary variable because c;l isnot in constraints (1)-(5) of the LRP. Intuitively, each cell c in X X  min m;c;l ¼ q À max pt;lthe subproblem is associated with a cost bc;l and a profitP l2V c:t2V ;c2Ct l2V t2V X m:c2Cm ;l2Lm m;c;l for each layer l, and the subproblem is to ¼ jV j  q À max pt;l ¼ jV j  q À p:minimize the net cost of each cell c for each layer l. t2V l2VTherefore, cell c needs to be selected for layer l in thesubproblem if the cost is less than the profit. In other words, The necessary condition can be proved similarly. u tP is one in the solution to the subproblem if bc;l is less thanc;l Although the second subproblem is NP-hard, the m:c2Cm ;l2Lm m;c;l . subproblem can be solved in a polynomial time if we The objective function of the second subproblem is the assume the number of technologies that can be used by eachsecond part of the objective function in the LRP: mobile host is a small constant, irrelevant to the input size X X X min m;c;l m;c;l : of the subproblem. Moreover, we assume that the number m2M c2Cm l2Lm of covering cells for each mobile host is also a small The second subproblem includes constraints (1)-(5) and constant. Based on the assumption, we solve the subpro-the integrality constraints for the variables. In the subpro- blem as follows: If the set of cells selected by a mobile host e eblem, each mobile host m needs to select a cell c for each m is C in any solution, then m must select the cell cC with m;lsubscribed layer l, where each choice is associated with a the minimum cost for each subscribed layer l, namely,nonnegative cost m;c;l . The subproblem is to minimize the etotal cost of the selected cells for all mobile hosts. cC ¼ arg min m;c;l : m;l In the following, we first prove that the second e c2Csubproblem is NP-hard: e C Therefore, the total cost m of the selected cells for eachTheorem 1. The second subproblem of the LRP is NP-hard. mobile host m is given as follows: Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 6. 280 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008 e X C ¼ m  e : and requires no modification on the current wireless m;cC ;l l2Lm m;l network infrastructures. Each mobile host m in algorithm LAGRANGE selects the To find the optimal solution, we consider every possible e e e cell c according to the cost m;c;l , the Lagrange multiplier, inset C, where C  Cm . Set C must follow two rules: First, the e the second subproblem. To reduce the total bandwidthnumber of technologies used in C is no more than tm . e consumption, we adjust the cost iteratively with theSecond, each technology has at most one cell in C: Let C subgradient algorithm [12] and the solutions to the twodenote the set of cells that induces the minimum cost, subproblems of the LRP. Let W ðÞ denote the objectivenamely, function of the LRP in Section 3.1, where & À Á e C ¼ arg min C : m  ¼ m;c;l ; 8m 2 M; 8c 2 Cm ; 8l 2 Lm : e C The subgradient corresponding to the optimal solution ofTherefore, for each subscribed layer l, let cC denote the cell m;l the LRP is denoted asselected by mobile host m; in other words, À Á & rW ðÞ ¼ @W ð Þ=@m;c;l ; 8m 2 M; 8c 2 Cm ; 8l 2 Lm ; m;c;l 1; ifc ¼ cC ; m;l 0; otherwise: where e The number of possible set C required to consider @W ð Þ  ¼ m;c;l À c;l : jtm jis OðjCm j Þ, and we need OðjLm jjtm jÞ comparisons to find @m;c;l ecC for all layers in Lm . Therefore, the time complexity of m;l The feasible solution to the CTSP obtained by algorithmour algorithm for the second subproblem of each member m LAGRANGE depends on the cost m;c;l , and the subgra-is OðjLm jjtm jjCm jjtm j Þ. We believe that the algorithm is dient @W ð Þ=@m;c;l indicates the direction of adjusting m;c;l to find the better feasible solution to the CTSP,suitable to be implemented in mobile hosts due to the 8m 2 M, 8c 2 Cm , 8l 2 Lm . Algorithm LAGRANGE in-following reasons: First, the algorithm is simple. For each creases or decreases the cost m;c;l at each iteration based eset C, the algorithm simply finds the cell with the minimum on the subgradient @W ð Þ=@m;c;l . Algorithm LAGRANGEcost, namely, the Lagrange multiplier, for each subscribed increases m;c;l when m;c;l À c;l is positive. In other words,layer. Second, since each mobile host tends to be covered by mobile host m selects cell c for layer l, but the cell is nota few cells and is able to access several wireless technol- selected in the first subproblem in this case. In contrast,ogies, we can regard jCm jjtm j as a small constant, and our algorithm LAGRANGE decreases m;c;l when m;c;l À c;l is negative. In other words, mobile host m does not selectalgorithm thereby can find the optimal solution to the cell c for layer l, but the cell is selected in the firstsecond subproblem in reasonable time. subproblem in this case. Since the variables to the first and second subproblems of We explain the adjustment of the costs in an intuitivethe LRP are mutually independent, the solution to the LRP way as follows: Although the solution to the first subpro-is just the combination of the solutions to the two blem may not find a feasible solution to the CTSP, the firstsubproblems. With the solution, algorithm LAGRANGE is subproblem provides an insight to select the cells a withable to find and improve the solution to the CTSP. low bandwidth cost. The first subproblem tends to select the cells that cover more mobile hosts to save the wireless3.2 Finding and Improving the Solution to the CTSP bandwidth because the cell tends to a own P larger profit,Although the first subproblem selects the cells with the where the profit of each cell c for layer l is m:c2Cm m;c;l .minimum net cost, the selected cells may not be feasible to Therefore, if mobile host m does not select cell c for layer lthe CTSP because each mobile host is not guaranteed to be but the cell is selected in the first subproblem, we decreasecovered by at least one selected cell. Therefore, our m;c;l such that the cell will more likely be chosen by m foralgorithm selects the cells to deliver each layer of a video layer l afterward. Therefore, the cost m;c;l , which is thestream according to the solution to the second subproblem. Lagrange multiplier in the LRP, plays an important role to eLet c;l be a binary variable that represents if our algorithm reduce the total bandwidth consumption. eselects cell c for the layer l of the video stream. Variable c;l Fig. 2 shows the details of our algorithm. Initially,as one if there exists at least one mobile host m selecting cell algorithm LAGRANGE assigns a unit cost to each cell forc in the second subproblem. In other words, variable m;c;l is each subscribed layer of each mobile host in step one, andone in the second subproblem. The advantage of algorithm each mobile host thereby can select any cell. Afterward, ourLAGRANGE is that the cell selection is performed only by algorithm iteratively reduces the total bandwidth cost of thethe mobile host and is transparent to the video multicasting selected cells and the wireless technologies. At eachmechanisms. In other words, each mobile host subscribes a iteration, our algorithm first finds the solutions to the firstlayer with the video multicasting mechanisms after it subproblem and the second subproblem in steps two andselects the cell and the wireless technology according to three, respectively. The algorithm then adjusts the cost ofour algorithm. We thereby require no modification on the each cell for each subscribed layer of each mobile host incurrent video multicasting mechanisms. In addition, algo- step four such that we can select the cells and the wirelessrithm LAGRANGE is implemented in only the mobile host technologies with lower bandwidth cost at the next Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 7. YANG AND CHEN: BANDWIDTH EFFICIENT VIDEO MULTICASTING IN MULTIRADIO MULTICELLULAR WIRELESS NETWORKS 281Fig. 2. Algorithm LAGRANGE.iteration. Algorithm LAGRANGE stops 1) when the In other words, the bandwidth cost of the cell is less thannumber of iterations  is larger than a threshold N, 2) when the profit, which is the sum of costs for the four mobileour algorithm can no longer adjust the cost, or 3) when the hosts. In the second subproblem, mobile hosts A and Bdifference of the total bandwidth cost of our solution and subscribe the two layers from cells Wi-Fi 1 and Wi-Fi 2, andthe lower bound on the total bandwidth cost of the optimal mobile hosts C and D subscribe the base layer from cellsolution is within a threshold . We enforce that the UMTS. Therefore, the two Wi-Fi cells need to deliver bothalgorithm can perform a limited number of iterations N to layers, and the UMTS cell needs to multicast the base layerfind the trade-off between the quality of the solution and of the video stream. In other words, variables WiÀFi1;1 ,ethe computational time. The parameter " in step five e e e e WiÀFi2;1 , WiÀFi1;2 , WiÀFi2;2 , and UMTS;1 are one in Fig. 3a,dominates the modification of the cost at each iteration in and the total bandwidth cost is 13.5.the subgradient algorithm. The solution with a larger " The first subproblem encourages the mobile hosts toimproves faster, but the solution with a smaller " converges select the UMTS cells for the layer since the cell is selectedto a better solution. In this paper, we thereby reduce " as in the first subproblem for Layer 1 of the video stream.the improvement of the selected cells becomes smaller. However, mobile hosts A and B do not select the UMTS cell Consider Fig. 3 as an illustrative example. The band- in the second subproblem. Therefore, algorithm LA-width cost of each UMTS and each Wi-Fi cell is 3.5 and 2.5 GRANGE reduces A;UMTS;1 and B;UMTS;1 to 0.8 accord-to deliver each layer of a video stream. Each mobile host can ingly. On the contrary, the two Wi-Fi cells are not selecteduse both wireless technologies. For simplicity, the video in the first subproblem, but mobile host A and B selects thestream here has only two layers, and the bandwidth two Wi-Fi cells in the second subproblem. Therefore,allocated for each UMTS and Wi-Fi cell can support one algorithm LAGRANGE increases A;WiÀFi1;1 , A;WiÀFi1;2 ,and two layers, respectively. The solid line connecting a B;WiÀFi2;1 , and B;WiÀFi2;2 to 1.2 accordingly, and Fig. 3bmobile host and a cell represents that the mobile host selects shows the new costs at the beginning of the secondthe cell for Layer 1, the base layer, of a video stream. The iteration. At the second iteration, both mobile hosts A andlong dash line indicates that the mobile host selects the cell B select the UMTS cell for Layer 1 of the video stream in thefor Layer 2, the enhancement layer, of the video stream. second subproblem because the cost for the UMTS cell isEach mobile host selects the cell for each subscribed layer smaller than the cost for the Wi-Fi cell. Similarly, bothaccording to the solution to the second subproblem of the mobile hosts A and B select cell Wi-Fi 1 for Layer 2 of theLRP. In the second subproblem of the LRP, each cell is e e video streams. Therefore, variables UMTS;1 and WiÀFi1;2 areassociated with a cost for each subscribed layer of each one in Fig. 3b, and the total bandwidth cost is 6, identical tomobile host, and the cost is shown beside each line in Fig. 3. the optimal solution in the example. Algorithm LA-In addition, the first subproblem of the LRP provides an GRANGE shows two advantages here: First, it selects theinsight for bandwidth-efficient cells, where each selected bandwidth-efficient wireless technology according to thecell selected is presented by a solid circle. bandwidth cost of each cell and the number of covered Initially, we set the cost of each cell as one. Therefore, mobile hosts. It uses the UMTS cell to serve all mobile hostsonly the UMTS cell is selected in the first subproblem for for Layer 1 of the video stream. Second, algorithmLayer 1 because the net cost of the cell is negative in Fig. 3a. LAGRANGE clusters the mobile hosts to reduce the Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 8. 282 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008Fig. 3. An example of algorithm LAGRANGE. (a) Iteration 1 ðcost ¼ 13:5Þ. (b) Iteration 2 before A moves ðcost ¼ 6Þ. (c) Iteration 2 after A movesðcost ¼ 8:5Þ. (d) Iteration 4 ðcost ¼ 6Þ.number of cells for each layer of the video stream. It selects Therefore, the costs A;WiÀFi2;2 and B;WiÀFi1;1 become 1.2 atcell Wi-Fi 1 to multicast Layer 2 of the video stream to the beginning of the third iteration. The third iteration onlymobile hosts A and B, instead of using the two Wi-Fi cells. increases the costs A;WiÀFi2;2 and B;WiÀFi1;1 to 1.4, and Figs. 3c and 3d show that algorithm LAGRANGE adapts Fig. 3d shows the new costs at the beginning of the fourthto the mobility of users. We assume that mobile host A iteration. At this iteration, mobile host B selects cell Wi-Fi 2moves to cell Wi-Fi 2 during the second iteration, and the for Layer 2 of the video stream since the cost B;WiÀFi2;2 istotal bandwidth cost increases to 8.5. The cost of cell Wi-Fi 2 smaller than B;WiÀFi1;2 . Therefore, we need only Wi-Fi 2 tofor Layer 2 of mobile host A is set as one in Fig. 3c. For deliver Layer 2 of the video stream. The total bandwidthLayer 1 of the video stream, each cell selected in the first cost is 6, which is identical to the optimal solution in thesubproblem is also selected in the second subproblem, and example. In addition, algorithm LAGRANGE stops at thiseach cell that is not selected in the first subproblem is not iteration because we can no longer increase or decrease theselected in the second subproblem either. Therefore, the cost of every cell.cost of each cell for Layer 1 of each mobile host remains the The above example shows that algorithm LAGRANGEsame. For Layer 2 of the video stream, since cells Wi-Fi 1 can reduce the bandwidth consumption to multicast a videoand Wi-Fi 2 are not selected in the first subproblem, stream in the heterogeneous wireless networks. Thisalgorithm LAGRANGE increases the cost of Wi-Fi 2 for example also indicates that algorithm LAGRANGE adaptsmobile host A and the cost of Wi-Fi 1 for mobile host B. to change of the locations of mobile hosts. Each mobile host Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 9. YANG AND CHEN: BANDWIDTH EFFICIENT VIDEO MULTICASTING IN MULTIRADIO MULTICELLULAR WIRELESS NETWORKS 283in the algorithm is able to individually select the cell for periodically initiated by sending a trigger message viaeach subscribed layer because each mobile host is able to multicast to all mobile hosts. The trigger message indicatessolve the corresponding two subproblems individually. the mode of the exchange of the costs and can be sent by theTherefore, algorithm LAGRANGE is suitable to be imple- video server. In the Unsuppress exchange of multipliers,mented in a distributed manner, and we design a protocol each mobile host obtains the profit of a covering cell after itfor the algorithm in Section 3.3. receives the costs from other mobile hosts in the cell. In the Suppress exchange of multipliers, each mobile host finds3.3 Protocol Design the profit of a covering cell with the profit stored in theThe proposed algorithm LAGRANGE can be implemented mobile host and the differences of the costs sent by otherin either a centralized or a distributed manner. In the mobile hosts in the cell. Afterward, each mobile host solvescentralized manner, the operator uses a server to find the the two subproblems of the LRP and selects the cell tosolution with our proposed algorithm. The centralized subscribe each layer of the video stream.approach requires no additional communication between Our protocol supports the mobility of users. When athe members or between the members and the base stations mobile host hands over to a new cell, it sends the cost that isto find the solution. However, it suffers from the scalability initially set as one to other mobile hosts in the new cellproblem because the number of servers is proportional to during the exchange of the costs. If the exchange of the coststhe number of video streams. Therefore, in this paper, we is in the Suppress mode, the mobile host also sends apropose a protocol to implement the algorithm in a negative cost to the old cell to reduce the total profit. Ourdistributed manner. The protocol can be implemented in protocol is resilient to the message loss. If any message withthe video subscription software in mobile hosts, and the the cost in the Suppress mode is lost, other mobile hosts cannetwork operator requires no server to provide the services. still obtain the correct profit of the cell after an UnsuppressIn this paper, we assume that the subscription of the layers exchange of the costs. If a mobile host keeps moving backis controlled in mobile hosts to ensure the scalability to and forth between adjacent cells, our algorithm can handlesupport more members and streams, just like the previous this situation by reducing the duration between twoworks. Therefore, our protocol informs each mobile host of iterations. The amount of time required by our algorithmthe identity of the cell and technology for each subscribed to find the best solution depends on the length of the timelayer, and we require no modification on the existing video between two iterations.multicast mechanisms. Our protocol supports the dynamic group membership. In the following, we present our protocol based on For a new member, it distributes and collects the Lagrangealgorithm LAGRANGE. We assume the wireless networks multipliers in the way as the member hands over into thesupport multicast between mobile hosts. Each mobile host cell. For a member that unsubscribes a video stream, thein our protocol periodically exchanges the costs with other behavior of the member in a cell is just like the way that themobile hosts in the same cell to solve the two subproblems member hands over out of the cell.of the LRP. Each mobile host selects the cell and technology Current video multicast mechanisms mainly focus onto subscribe each layer of the video stream from the adaptively adding and dropping a layer according to thesolutions of the two subproblems. Each cell needs to deliver available bandwidth to avoid quality fluctuation. Thea layer of the video stream if at least one mobile host in the current mechanisms adopt buffering techniques to synchro-cell subscribes the layer from the cell. nize the playback of each layer at mobile hosts. Each mobile To reduce the protocol overhead, we divide the host in the mechanisms uses a state machine to ensure thatexchange of the costs into two modes: Unsuppress and the addition or drop of a layer is in the progressive manner.Suppress. Our protocol inserts multiple Suppress ex- The state machine also enforces that each mobile host mustchanges between two Unsuppress exchanges of the costs. first subscribe the base layer when it would like to receive aIn the Unsuppress mode, each mobile host sends the costs video stream. In addition, the state machine includes timersvia multicast to other mobile hosts in each covering cell. In and hysteresis algorithms to avoid the ping-pong effect ofthe Suppress mode, however, each mobile host suppresses the quality fluctuation between layers. In contrast, oursending a cost to other mobile hosts if the difference of the approach is orthogonal to them and needs to adopt one ofcurrent cost and the cost that was sent to other mobile them to subscribe a video stream.hosts is within a threshold. Each mobile host m in ourprotocol stores the following information for each coveringcell c: 1) The cost that was sent to other mobile hosts for 4 EXPERIMENTAL STUDIESeach layer l in Lm . 2) The profit of the cell for each layer l In this section, we present our simulation results. As wein the first subproblem of the LRP. The profit is the sum of know, there is no related algorithm for the CTSP in thethe total costs of all mobile hosts in the cell, namely, previous works. Therefore, we compare algorithm LA-P m:c2Cm ;l2Lm m;c;l . A mobile host obtains the profit by GRANGE with two other algorithms that can represent theexchanging the Lagrange multipliers with other mobile reasonable user behaviors. In the first algorithm RAND,hosts in the cell. each mobile host randomly selects a cell to subscribe all Each mobile host needs to access each covering cell only layers of a video stream. In the second algorithm LOCAL,during the exchange of multipliers. In other times, each each mobile host individually selects the wireless technol-mobile host accesses only the selected cell to receive the ogy with the minimum bandwidth cost to subscribe allvideo stream. In our protocol, the exchange of multipliers is layers of the video stream because the mobile host tends to Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 10. 284 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008Fig. 4. Simulation results of homogeneous wireless networks. (a) The bandwidth cost to deliver the video stream. (b) The bandwidth cost ofCONTROL.spend the least monetary cost in this case. The mobile host obtained by algorithm LAGRANGE adopts the SUPPRESSselects the cell with the minimum distance to the base exchange of the costs.station if there are multiple cells with the technology In our simulation, we adopt the probability-basedcovering the mobile host. In addition to RAND and mobility model widely used in the previous works. Initially,LOCAL, we compare our solution with the optimal solution we distribute all mobile hosts uniformly at random in theOPT obtained by CPLEX [27] with our ILP formulation in networks in a 10 km  10 km service area. Afterward, eachsmall networks. We also compare the total bandwidth cost mobile host randomly decides to stay or move at eachof our solution with the lower bound LB on the total minute. If the mobile host decides to move, it randomlybandwidth cost of the optimal solution in large networks, chooses the direction and the speed from 0 to 1 km perwhere we obtain the lower bound by solving the LRP in minute. Our simulation adopts the cellular network infra-Section 3. structure, which has been widely used in the experiments of To test the performance of our algorithm in different related works. In addition, we decide the locations of Wi-Fiscenarios, we change the following parameters: cells randomly because we assume that Wi-Fi cells need to cover only hotspot areas and, as we know, there is no number of mobile hosts that subscribe the video 1. related work on the distribution of hotspot areas. The stream to consider the scalability of our algorithm mobility model in this paper is very similar to the Random and protocol, Walk model that has been widely adopted in the literature. 2. number of wireless technologies that can be used by However, we assume that mobile hosts sometimes can stop each mobile host to consider the heterogeneity of for a while in this paper to model the behaviors of users, pedestrians or vehicles in a city. For algorithm LA- 3. the bandwidth cost of each wireless technology, and GRANGE, we run one iteration every 30 seconds. The 4. transmission range of a base station. simulation time takes 1,000 minutes.For each wireless technology, the size of the overlapping We first compare the solutions obtained by our algo-area of adjacent cells changes when the transmission range rithms with the optimal solutions obtained by CPLEX withof a base station is different. We measure 100 samples in our ILP formulation. Because solving large ILP problems iseach scenario with 95 percent confidence interval. computational infeasible, here, we simulate the scenario The performance metrics in our simulation are listed as that the video stream is not layer encoded, and the networkfollows: 1) Bandwidth cost induced by each algorithm. We operators own only one wireless technology. The networkmeasure the total bandwidth cost and the bandwidth cost has 144 Wi-Fi cells, where the bandwidth cost of each cell isof each wireless technology to deliver a video stream. three. Fig. 4a presents the bandwidth costs of algorithms2) Control overhead of our protocol, denoted as CON- LAGRANGE, RAND, LOCAL, and OPT. Fig. 4a shows thatTROL. We measure the bandwidth cost of the cells to LAGRANGE outperforms RAND and LOCAL becauseexchange the costs in order to compare CONTROL with the LAGRANGE clusters the mobile hosts to reduce thebandwidth costs to send the video stream. The number of number of cells to multicast the video stream. Mobile hostscontrol messages in our protocol is proportional to the in LAGRANGE jointly reduce the total bandwidth con-number of cells to exchange the costs and the number of sumption by exchanging the costs. On the contrary, eachmobile hosts in a cell. In our simulation, we compare the mobile host in RAND and LOCAL selects the cellsoverhead of our protocol in only Unsuppress mode, independently. Fig. 4b shows that our protocol in Suppressdenoted as UNSUPPRESS, with the overhead of our mode can reduce the number of cells to exchange the costs.protocol in both Unsuppress and Suppress modes, denoted Therefore, the above results indicate that our algorithm andSUPPRESS. In the SUPPRESS exchange of the costs, there protocol can reduce the total bandwidth consumption evenare five Suppress exchanges of the costs between two when the video stream is not layer encoded, and theUnsuppress exchanges. In our simulation, the solution network operators own only one wireless technology. Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 11. YANG AND CHEN: BANDWIDTH EFFICIENT VIDEO MULTICASTING IN MULTIRADIO MULTICELLULAR WIRELESS NETWORKS 285Fig. 5. Simulation results of heterogeneous wireless networks. (a) The total bandwidth cost of both Wi-Fi cells and UMTS cells. (b) The bandwidthcost of CONTROL. (c) The bandwidth cost of Wi-Fi cells. (d) The bandwidth cost of UMTS cells. Our protocol is based on the proposed algorithm UMTS cell induces less bandwidth cost than multiple Wi-FiLAGRANGE, which needs to exchange the Lagrange cells as the number of mobile hosts increases. Algorithmmultipliers in not only the cells that deliver the stream, RAND uses more UMTS cells than LAGRANGE becausebut also the cells that cover the mobile hosts but do not each mobile host selects the cell individually, regardless ofmulticast the stream. The mobile hosts need to exchange the the selections of other mobile hosts. Although algorithmmultipliers in the latter cells to reduce the bandwidth LAGRANGE uses more UMTS cells than LOCAL, the totalconsumption in a distributed manner. Therefore, more cells bandwidth cost of LAGRANGE is much less than the cost ofare counted in the results of CONTROL. However, each LOCAL, as shown in Fig. 5a. The reason is thatmobile host only accesses the latter cells during the LAGRANGE has the abilities to cluster the mobile hostsexchange of multipliers. and to select the proper wireless technology according the In the following, we test our algorithm in the hetero- number and locations of mobile hosts.geneous wireless networks. The network has 144 Wi-Fi cells We then test our algorithm when the number of wirelessand four UMTS cells, where the bandwidth cost of each Wi- technologies that can be used by each mobile host isFi cell and UMTS cell is 3 and 15, respectively, for each layer different. Some mobile hosts would like to improve theof the video stream. Fig. 5a presents the total bandwidth video quality by subscribing more layers with both wirelesscost to multicast the video stream. Algorithm LOCAL is technologies, whereas others may concern power consump-slightly better than RAND because the mobile hosts in the tion and thereby subscribe a limited number of layers withLOCAL prefer to use the Wi-Fi cells. However, mobile hosts a single wireless technology. In the following, the video hasin RAND waste the network bandwidth because multiple four layers here, and the bandwidth allocated for eachmobile hosts in the same cell may choose the different UMTS and Wi-Fi cell can support two and four layers. Wewireless technologies even when only one technology is change the proportion of mobile hosts that can use bothrequired here. In contrast, algorithm LAGRANGE outper- wireless technologies in the networks. Fig. 6a shows thatforms both RAND and LOCAL, and the solutions obtained LAGRANGE uses more UMTS cells as more mobile hostsby LAGRANGE and LB are very close. can use both technologies. However, Fig. 6b shows that the Figs. 5c and 5d compare the total bandwidth consump- total bandwidth cost decreases as more mobile hosts cantion in each wireless network. Algorithms RAND and use both technologies. The reason is that more mobile hostsLOCAL select more Wi-Fi cells as the number of mobile subscribe Layers 1 and 2 of the video stream from thehosts increases. In contrast, algorithm LAGRANGE selects UMTS cells, and fewer Wi-Fi cells are required to multicastfewer Wi-Fi cells but more UMTS cells because a single the two layers. Therefore, these cells save the bandwidth of Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 12. 286 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008Fig. 6. Simulation results of different percentages of mobile hosts that can use the both wireless technologies. (a) The bandwidth cost of UMTS cells.(b) The total bandwidth cost of both Wi-Fi and UMTS cells.Fig. 7. Simulation results of different bandwidth costs assigned to each UMTS cells. (a) The number of used UMTS cells. (b) The number ofhandovers for each mobile host.Fig. 8. Simulation results of different radii of a cell with the ratio respect to the original radius. (a) The total bandwidth cost of both Wi-Fi and UMTScells. (b) The number of handovers for each mobile host.the two layers and have the ability to multicast more layers the cell is congested. In addition, Fig. 7b shows that a largerto the mobile hosts to improve the quality of the video cost for UMTS cells also leads to fewer handovers for eachstream. mobile host. The network operators can change the bandwidth cost of Fig. 8a shows the total bandwidth cost when we changeeach cell to adjust the distribution of traffic. Fig. 7a shows the radius of a cell with the ratio with respect to the originalthe total number of cells to multicast the video stream with radius. As the radius increases, each mobile host is covereddifferent costs to the UMTS cells. With a larger cost by more cells, and more mobile hosts can be clusteredassigned to each UMTS cell, algorithm LAGRANGE uses together to use fewer cells. Therefore, algorithm LA-fewer UMTS cells in Fig. 7a. Therefore, the network GRANGE requires less network bandwidth to multicastoperators can increase the bandwidth cost of a cell when the video stream. However, our algorithm in this case Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 13. YANG AND CHEN: BANDWIDTH EFFICIENT VIDEO MULTICASTING IN MULTIRADIO MULTICELLULAR WIRELESS NETWORKS 287 solution at the 43rd iteration, which is better than the one at the 14th iteration. Fig. 9b shows the transient behavior of our protocol based on the proposed algorithm. There are nine Suppress exchanges of the costs between two Unsuppress exchanges of the costs in our simulation. We change the threshold in the Suppress mode of our protocol. Each mobile host suppresses sending the cost if the difference of the current cost and the previous sent cost is within the threshold. Here, we let all mobile hosts move at iteration 30 to show the effect of mobility in our algorithm. Algorithm LA- GRANGE iteratively reduces the total bandwidth consump- tion, as shown in Fig. 9b. After the mobile hosts move, algorithm LAGRANGE consumes more bandwidth. How- ever, algorithm LAGRANGE reduces the total bandwidth consumption iteratively after the mobile hosts exchange the costs. Fig. 9c shows the control overhead of our protocol. Our protocol induces more overhead as the threshold increases in Fig. 9c. However, our protocol in this case also converges faster toward the optimal solution, as shown in Fig. 9b. 5 CONCLUSION In this paper, we propose a new mechanism to select the cells and the wireless technologies for layer-encoded video multicasting in heterogeneous wireless networks. Each mobile host in our mechanism can select a different cell to subscribe each layer of the video stream, and each cell can multicast only a subset of layers of the video stream to reduce the bandwidth consumption. We formulate CTSP in the heterogeneous wireless networks as an optimization problem. We use ILP to model the problem. The network operators can use the ILP formulation to find the optimal solutions for network planning. We show that the problem is NP-hard and design an algorithm LAGRANGE, which is based on Lagrangean relaxation on our ILP formulation. Algorithm LAGRANGE iteratively converges toward the optimal solutions and can be implemented in the distrib- uted manner in only the mobile hosts. Our mechanism requires no change of the current video multicastingFig. 9. Thansient behavior of algorithm LAGRANGE. (a) The bandwidth mechanisms and the current wireless network infrastruc-cost of LAGRANGE. (b) The bandwidth cost of LAGRANGE. (c) Thebandwidth cost of CONTROL. tures. Our algorithm is adaptive to the change in the subscribers at each layer and the change of the location ofrequires more iterations to find the solution because each each mobile host. Our algorithm is flexible such thatmobile host has more candidate cells for each subscribed network operators can balance the load of the wireless cellslayer. Therefore, each mobile host needs more handovers in with the same or different wireless technologies.this case, as shown in Fig. 8b. Fig. 9a shows the transient behavior of our algorithm. ACKNOWLEDGMENTSWith a large ", that is, " ¼ 10 in this case, the bandwidth cost The authors would like to thank the anonymous referees forconverges faster to 400 with only 14 iterations, whereas a their helpful comments to improve this paper.small ", that is, " ¼ 1 in this case, requires 29 iterations.However, our algorithm with a small " can obtain a bettersolution that approaches the lower bound. Our algorithm REFERENCESsometimes obtains a solution slightly worse than the [1] W. Kumwilaisak et al., “A Cross-Layer Quality-of-Service Map-solutions of the nearby iterations. The logic behind ping Architecture for Video Delivery in Wireless Networks,” IEEEsearching toward a locally worse direction is to avoid being J. Selected Areas in Comm., vol. 21, no. 10, pp. 1685-1698, Dec. 2003. [2] M. Chen and A. Zakhor, “Rate Control for Streaming Video overtrapped in a locally optimal solution in order to find a better Wireless,” IEEE Wireless Comm., vol. 12, no. 4, pp. 32-41, Aug.solution later. Our algorithm, a large " in Fig. 9a, obtains a 2005. Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.
  • 14. 288 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 7, NO. 2, FEBRUARY 2008[3] Y. Liang and B. Girod, “Network-Adaptive Low-Latency Video De-Nian Yang received the BS and PhD Communication over Best-Effort Networks,” IEEE Trans. Circuits degrees from the Department of Electrical and Systems for Video Technology, vol. 16, no. 1, pp. 72-81, Jan. 2006. Engineering, National Taiwan University, Taipei,[4] W. Li, “Overview of Fine Granularity Scalability in MPEG-4 Video Taiwan, in 1999 and 2004, respectively. He is Standard,” IEEE Trans. Circuits and Systems for Video Technology, currently a postdoctoral researcher in military vol. 11, no. 3, pp. 301-317, Mar. 2001. service in the Department of Electrical Engineer-[5] J. Ohm, “Advances in Scalable Video Coding,” Proc. IEEE, vol. 93, ing, National Taiwan University. His research no. 1, pp. 42-56, Jan. 2005. interests include network planning, multicasting,[6] J.-L. Huang, W.-C. Peng, and M.-S. Chen, “SOM: Dynamic Push- and Quality of Service in wireless networks. He Pull Channel Allocation Framework for Mobile Data Broad- is a member of the IEEE. casting,” IEEE Trans. Mobile Computing, vol. 5, no. 8, pp. 974-990, Aug. 2006.[7] S. McCanne, V. Jacobson, and M. Vetterli, “Receiver-Driven Ming-Syan Chen received the BS degree in Layered Multicast,” Proc. ACM SIGCOMM, vol. 26, no. 4, electrical engineering from the National Taiwan pp. 117-130, 1996. University, Taipei, Taiwan, and the MS and PhD[8] B. Vickers, C. Albuquerque, and T. Suda, “Source-Adaptive degrees in computer, information, and control Multilayered Multicast Algorithms for Real-Time Video Distribu- engineering from the University of Michigan, Ann tion,” IEEE/ACM Trans. Networking, vol. 8, no. 6, pp. 720-733, Dec. Arbor, Michigan, in 1985 and 1988, respectively. 2000. He is currently a professor and the chairman of[9] T. Kim and M. Ammar, “Optimal Quality Adaptation for MPEG-4 the Graduate Institute of Communication En- Fine-Grained Scalable Video,” Proc. IEEE INFOCOM, vol. 1, gineering, a professor in the Electrical Engineer- pp. 641-651, ing Department, and also a professor in the[10] J. Liu, B. Li, Y. Hou, and I. Chlamtac, “On Optimal Layering and Computer Science and Information Engineering Department, National Bandwidth Allocation for Multisession Video Broadcasting,” IEEE Taiwan University, Taipei, Taiwan. He was a research staff member at Trans. Wireless Comm., vol. 3, no. 2, pp. 656-667, Mar. 2004. the IBM Thomas J. Watson Research Center, Yorktown Heights, New[11] S. Zhao, Z. Xiong, and X. Wang, “Optimal Resource Allocation for York, from 1988 to 1996. His research interests include database Wireless Video over CDMA Networks,” IEEE Trans. Mobile systems, data mining, mobile computing systems, and multimedia Computing, vol. 4, no. 1, pp. 56-67, Jan./Feb. 2005. networking, and he has published more than 180 papers in his research[12] G.L. Nemhauser and L.A. Wosley, “Integer and Combinatorial areas. In addition to serving as a program committee member in many Optimization,” Wiley-Interscience Series in Discrete Mathematics and conferences, he served as an associate editor of the IEEE Transactions Optimization, 1999. on Knowledge and Data Engineering from 1997 to 2001, is currently on[13] X. Yong, D. Harrison, S. Kalyanaraman, K. Ramachandran, and A. the editorial board of the Very Large Databases Journal, Knowledge and Venkatesan, “Accumulation-Based Congestion Control,” IEEE/ Information Systems Journal, Journal of Information Science and ACM Trans. Networking, vol. 13, no. 1, pp. 69-80, Feb. 2005. Engineering, and the International Journal of Electrical Engineering,[14] I. Paschalidis and Y. Liu, “Pricing in Multiservice Loss Networks: and is also a guest coeditor for the IEEE Transactions on Knowledge Static Pricing, Asymptotic Optimality and Demand Substitution and Data Engineering on a special issue for data mining in December Effects,” IEEE/ACM Trans. Networking, vol. 10, no. 3, pp. 425-438, 1996. He was a distinguished visitor of the IEEE Computer Society for June 2002. Asia-Pacific from 1998 to 2000. He served as the program chair of the[15] C. Beard and V. Frost, “Prioritized Resource Allocation for Pacific Area Knowledge Discovery and Data Mining (PAKDD ’02), Stressed Networks,” IEEE/ACM Trans. Networking, vol. 9, no. 5, program vice chair of the IEEE International Conference on Distributed pp. 618-633, Oct. 2001. Computing Systems (ICDCS) 2005 and the International Conference on[16] M. Chiang, “Balancing Transport and Physical Layers in Wireless Parallel Processing (ICPP) 2003, program vice chair of Very Large Multihop Networks: Jointly Optimal Congestion Control and Databases Conference (VLDB ’02), and also the general chair and Power Control,” IEEE J. Selected Areas in Comm., vol. 23, no. 1, program chair of several other conferences. He was a keynote speaker pp. 104-116, Jan. 2005. on Web data mining at the International Computer Congress in Hong[17] Y. Zhang, O. Yang, and H. Liu, “A Lagrangean Relaxation and Kong in 1999, a tutorial speaker on Web data mining at the International Subgradient Framework for the Routing and Wavelength Assign- Conference on Database Systems for Advanced Applications (DASFAA ment Problem in WDM networks,” IEEE J. Selected Areas in Comm., ’99) and on parallel databases at the 11th IEEE International vol. 22, no. 9, pp. 1752-1765, Nov. 2004. Conference on Data Engineering (ICDE) in 1995. He holds or has[18] E.W.M. Wong, A.K.M. Chan, and T.-S.P. Yum, “Analysis of applied for 18 US patents and seven ROC patents in the areas of data Rerouting in Circuit-Switched Networks,” IEEE/ACM Trans. mining, Web applications, interactive video playout, video server design, Networking, vol. 8, no. 3, pp. 419-427, June 2000. and concurrency and coherency control protocols. He is a recipient of[19] K.-C. Lee and V.O.K. Li, “A Wavelength Rerouting Algorithm in the National Science Council (NSC) Distinguished Research Award and Wide-Area All-Optical Networks,” IEEE J. Lightwave Technology, the K.-T. Li Research Penetration Award for his research work and the vol. 14, no. 6, pp. 1218-1229, June 1996. Outstanding Innovation Award from IBM Corporate for his contribution to[20] A. Donner, M. Berioli, and M. Werner, “MPLS-Based Satellite a major database product. He also received numerous awards for his Constellation Networks,” IEEE J. Selected Areas in Comm., vol. 22, research, teaching, inventions, and patent applications. He is a fellow of no. 3, pp. 438-448, Apr. 2004. the IEEE and the ACM.[21] K. Pahlavan et al., “Handoff in Hybrid Mobile Data Networks,” IEEE Personal Comm., vol. 7, no. 2, pp. 34-47, Apr. 2000.[22] J. McNair, I.F. Akyildiz, and M. Bender, “An Inter-System Handoff Technique for the IMT-2000 System,” Proc. IEEE . For more information on this or any other computing topic, INFOCOM, vol. 1, pp. 208-216, 2000. please visit our Digital Library at www.computer.org/publications/dlib.[23] J. McNair and F. Zhu, “Vertical Handoffs in Fourth-Generation Multinetwork Environments,” IEEE Wireless Comm., vol. 11, no. 3, pp. 8-15, June 2004.[24] Y. Pan, M. Lee, J.B. Kim, and T. Suda, “An End-to-End Multipath Smooth Handoff Scheme for Stream Media,” IEEE J. Selected Areas in Comm., vol. 22, no. 4, pp. 653-663, May 2004.[25] B. Fortz and M. Thorup, “Optimizing OSPF/IS-IS Weights in a Changing World,” IEEE J. Selected Areas in Comm., vol. 20, no. 4, pp. 756-767, May 2002.[26] M.R. Gary and D.S. Johnson, Computer and Intractability: A Guide to the Theory of NP-Hardness. W.H. Freeman, 1979.[27] CPLEX Mathematical Programming Optimizer, http://www.ilog. com/products/cplex/, 2007. Authorized licensed use limited to: GOVERNMENT COLLEGE OF TECHNOLOGY. Downloaded on January 22, 2009 at 02:23 from IEEE Xplore. Restrictions apply.