A Broadcast-VOD Protocol in an Integrated Wireless Mobile Network

全文

(1)A Broadcast-VOD Protocol in an Integrated Wireless Mobile Network Yuh-Shyan Chen and Yuan-Chun Lin Department of Statistics, National Taipei University Taipei, 10433, Taiwan yschen@mail.ntpu.edu.tw. Abstract In this paper, we investigate a broadcast-VOD problem in an integrated wireless mobile network. Existing wireless mobile networks can be classified into single-hop (such as GSM system) and multi-hop mobile networks (such as Ad-hoc network). A single-hop wireless mobile network is characterized by each base station (or BS) connecting to other BS by a wired network, but each mobile host (or MH) communicating with its own BS within one-hop transmission radius. Our proposed wireless mobile network is an integration of single-hop and multi-hop mobile networks. The integrated network is characterized by each BS connecting to other BS by a wired network, but each MH communicating with a BS within multi-hop transmission steps. In our integrated network, BSs are connected by a 2-D tori interconnection network. In addition, an important application, called as broadcast-VOD protocol, is presented in such integrated network. Therefore, two contributions are presented in this paper: (a) a new integrated wireless mobile network is proposed, (b) a fast broadcastVOD protocol is introduced in such new network. In our strategy, VOD-data are well scheduled and broadcasted among 2-D tori-BSs such that the average waiting time is minimum. In each cell area, a VOD (Video On-Demand) is dynamically requested by any MH at the different time to on-line construct a multicast tree, while the root node of multicast tree is a BS. This guarantees that each MH can receive the VOD-segment from its corresponding BS within minimal waiting time. Performance analysis verifies that our proposed scheme outperforms existing schemes.. 1. Introduction. A video system is consisting of a video-server with a high capability of disk-array, a transport network (ATM or cablemodem), and the end user. Up to date, there are many VOD research results in the wired-network [1][6][9][11][17]. For instance, an optimal video placement strategy is proposed [9] in the high-speed ATM network. In addition, Su and Wang [21] proposed an on-demand multicast routing Scheme. Their scheme allows the destinations accessing the same multicast stream at different time by using a buffering technique to reduce the communication bandwidth. Existing research system are mainly based on either highspeed network (ATM, Ethernet) or cable-TV network [1] [6] [9][11][17]. The requisition for bandwidth and quality does not allow the efficient implementation of such demanding services as VOD over the wireless mobile network. It is observed that a wireless communication usually requires digital data to be protected against channel error, as the communication link tends to be less reliable than in a Contact author: Yuh-Shyan Chen. This work was supported by the National Science Council, R.O.C., under Contract NSC89-2218-E-305-003.. wired-network. However, recently, there are many VOD research results in the wireless network [8][16][23][26]. Initially, Meng [23] proposed a wireless video system with low-power consumption and high computation performance which is based on their proposed elegant compression algorithm. Recently, Xu et al. [26] have proposed a QoSdirected error control scheme of video multicast in the wireless network. Note that their scheme has high error recovery rate for the video frames and with excellent scalability. In addition, Zheng et al. [8] designed a QoS-aware mobile video communication strategy. This strategy utilizes a new video-frame compressed technique to satisfy the QoS requirements. More recently, Davies et al. [16] developed a supporting adaptive video applications in the mobile environment. As a conclusion, we observe that existing VODin-wireless researchers focused on how to build a QoS transmission in the wireless network. This paper is motivated by developing a new mobile network model. In such new model, we investigate a fast broadcast-VOD problem in the new wireless network. Existing wireless mobile networks are classified into cellular network architecture (or single-hop wireless network with a fixed infrastructure, such as GSM system) and multihop wireless mobile networks (without a fixed infrastructure, such as Ad-hoc network). Mobile Ad-hoc networks (MANET) [10] [24] [7] [12] [15] [18] [3] consists of wireless hosts that communicate with each other, in the absence of a fixed infrastructure. A cellular network architecture with a fixed infrastructure is normally constituted by a set of base stations, while each base station communicates with a mobile host within a single-hop communication radius. A cellular network architecture is characterized by each base station connecting to other base station by means of the wired-network, but each mobile host communicates with its own base station within one-hop transmission radius. Due to considerations such as radio power limitations, power consumption, and channel utilization, a mobile host may not be able to communicate directly with other hosts in a single-hop fashion. A multi-hop scenario occurs, where the packets sent by the source host are retransmitted by several intermediate hosts before reaching the destination host. In MANET, host mobility can cause frequent unpredictable topology changes, thus design protocols more complicated than traditional single-hop wireless mobile network. Unfortunately, it is existing fewer applications in the multi-hop wireless network. Therefore, this paper is to present an integrated wireless network model, which is possessing natural advantage of both wireless networks. A related work is that Qiao et al. [20] proposed an integrated cellular and ad-hoc relay system. Qiao et al. [20] propose to integrate the cellular infrastructure with modern ad-hoc relaying technologies.

(2) to achieve dynamic load balancing among different cells in a cost-effective way. The major contribution of this paper is to propose an integrated wireless mobile network. Our integrated network is characterized by each base station connecting to other base station by a wired network, but each mobile host communicates with a BS within multi-hop transmission steps. Observe that, in our integrated network, BSs are connected by a 2-D tori interconnection network, which is called as the 2-D tori-BSs. One important advantage of our new network is to build a wireless network infrastructure by using less number of BSs. Our proposed scheme is suitable to build a wireless network infrastructure with low hardware cost. That is, there only needs fewer number of base stations in the same covering area, but each mobile host must communicate to a base station by the multi-hop fashion. One important technique in single-hop wireless network is the video-on-demand problem, such that mobile user can retrieve video data through base station [17][11][9][1][6]. Some applications using the above technique is the distance learning and video conference, etc. In this paper, we will propose a fast broadcast-VOD scheme in our integrated network to handle the video-on-demand (VOD) problem. Therefore, two contributions are presented in this paper; (a) a new integrated wireless mobile network is proposed, (b) a fast broadcast-VOD protocol is introduced in such new network. In our strategy, VOD-data are well scheduled and broadcasting among 2-D tori-BSs such that the average waiting time is minimum. In each cell area, a VOD (Video On-Demand) is dynamically requested by any MH at the different time to on-line construct a multicast tree, while the root node of multicast tree is a BS. This guarantees that each MH can receive the VOD-segment from its corresponding BS within minimal waiting time. Performance analysis verifies that our proposed scheme outperforms existing schemes. The rest of the paper is organized as follows. Section 2 introduces the new wireless network model. Our broadcastVOD protocol is presented in Section 3. Section 4 illustrates performance analysis. Section 5 concludes this paper.. 2. The Integrated Wireless Network Model. We now formally define our integrated wireless mobile network. Our integrated network is characterized by each BS connecting to other BS by a wired network, but each MH communicating with a BS within multi-hop transmission steps. A wireless network can be formally represented as G = (V; C ), where V denotes the BS set and C specifies the wired-network connectivity. Each BS contains a separate router to handle its communication tasks. In this paper, we consider G as a 2-dimensional torus Tn1 n2 with n1 n2 BSs [25]. Each BS is denoted as B i; j , 1 i n1 , 1 j n2 and Bi1 ;i2 has an edge connected B (i1 1) modn1 ;i2 along dimension one and an edge to B i1 ;(i2 1) modn2 along dimension two. Each edge is considered consisting of two directed communication links pointing in opposite directions. An example of Bi; j , 1 i 5 and 1 j 5; is given in Fig. 1. Each cell area is defined as a MH’s moving area which bounds by four neighboring BSs as illustrated in Fig. 1, and each MH can connect to four neighboring BSs by the multi-hop fashion. In our integrated network, BSs are connected by a 2D tori interconnection network [25]. For simplicity, such. Figure 1: An example of B i; j , where 1 i 5 and 1 j 5;. structure is denoted as 2-D tori-BSs. We make the following observation. The main advantage of our integrated network is to reduce the number of BSs. Our integrated wireless mobile network is used to support a wireless network infrastructure using fewer number of BSs. The proposed broadcast-VOD protocol is divided into two parts. 1. The Broadcast Protocol – The original video data is partitioned into data subsegments, which are located on BSs, scheduled and broadcasting on the 2-D toriBSs. 2. The VOD Protocol – Each MH performs the VOD protocol to receive the first data subsegment with minimal waiting time.. 3 The Proposed Broadcast Protocol Two operations of broadcasting protocol in 2-D tori-BSs are introduced.. . Data Assignment Operation – It describes the data initialization of the video segments into the 2-D toriBSs. Data Broadcast Operation – It denotes the data scheduling and broadcasting operations in the 2-D toriBSs.. In the following, we will describe these operations in detail.. 3.1 Data Assignment Operation. Consider that there is n n 2-D tori-BSs. Given a video, denoted as S, which is divided into 2n 2 data subsegments, so that S is equally split into fS 1 ; S2 ; S3 ; : : : ; S2n2 g. It reflects the fact that each BS keeps two data subsegment at the same time. In other words, we assume that each BS just.

(3) Figure 2: An example of video-partition scheme.. Figure 4: Data assigement of G e .. Figure 3: Data assigement of G o . only keeps two data subsegments at any time. This indicates that our scheme has the advantage of low memory cost. An instance is shown in Fig. 2, if we assume that n = 5 and a video is divided into 50 data subsegments. In the following, we will describe how to allocate the fS1; S2 ; S3; : : : ; S2n2 g into n n 2-D tori-BSs. We begin by classified the fS1 ; S2 ; S3 ; : : : ; S2n2 g into two groups. First group, denoted as G o , consists of odd number of datasubsegments fS1 ; S3 ; S5 ; : : : ; S2n2 ;1 g. Second group, denoted as Ge , consists of even number of data-subsegments fS2; S4 ; S6; : : : ; S2n2 g. Since the data allocation rule of G o and Ge is similar, therefore, we mainly focus on describing the constructing rule of G o : Recall above instance as illustrated in Fig. 2, a video with 50 data subsegments is split into fS1 ; S3 ; S5 ; : : : ; S49 g and fS2 ; S4 ; S6 ; : : : ; S50 g. Given Go = fS1 ; S3 ; S5 ; : : : ; S2n2 ;1 g, Go is again partitioned into two subgroups. The first subgroup, denoted as Go (1); where Go (1) = fS1 ; S3 ; S5 ; : : : ; S2n;1 g, and all of the remaindering data subsegments fS 2n+1 ; S2n+3 ; S2n+5 ; : : : ; S2n2 ;1 g are collected into second subgroup, which denoted as Go (2). The allocation of G o (1) and Go (2) into n2 2-D tori-BSs is formally given below.. Figure 5: The combing result of G o and Ge .. G1: Each element of G o (1) = fS1 ; S3 ; S5 ; : : : ; S2n;1 g is sequentially put into B 1;1 ; B2;n ; B3;1 ; B4;n ; : : : ; and Bn;1 . - For instance, consider a B i; j , 1 i 5, 1 j 5 as shown in Fig. 3(a), every element of G o (1) =. Figure 6: Examples of (a)(d) row-shift operations and (b)(c) column-shift operations..

(4) 3.2 Data Broadcast Operation Before describing the data broadcast operation of G o and Ge on n2 2-D tori-BSs, we now define two kinds of parallel data shifting operations.. Figure 7: An example of data retrieving operation.. Row-Shift Operation – For each row, a left-shift operation is performed in parallel if the row is an odd row. Otherwise, a right-shift operation is executed in parallel. - An example of the row-shift operations is illustrated in Fig. 6(a) and Fig. 6(d).. . Column-Shift Operation – A up-shift operation is performed in parallel for each column. - An instance of the column-shift operation is operated in Fig. 6(b) and Fig. 6(c).. We now introduce the data broadcast operation in the 2-D tori-BSs. The data broadcast operation is achieved by executing the data transmission of G 0 and Ge simultaneously.. The data transmission of G o is achieved by repeatedly performing row-shift and column-shift operations. On the contrary, the data transmission of G e is achieved by repeatedly performing column-shift and row-shift operations. A fully example is shown in Fig. 6 to express the data transmission operation of G o and Ge : Based on the rule of our data assignment and transmission operations, each BS B has two data subsegments S o and Se at the same time, where S o 2 Go and Se 2 Ge . Assume that B will receive S o0 and Se0 from row-shift and column-shift operations, where S o0 2 Go and Se0 2 Ge : Two possible data retrieving operations are given below.. Figure 8: An example of data broadcast operation.. fS1 S3 S5 S7 S9 g is put into B1 1 ;. ;. ;. ;. ;. ;. B2;4 ; B3;1 ; B4;4 ; and. . If now is using S o , then Se0 will be used on next time unit.. . If now is using S e , then So0 will be used on next time unit.. B5;0 in order.. G2: Each. element. fS2n. of. G o (2). =. is sequentially +1 ; S2n+3 ; S2n+5 ; : : : ; S2n2 ;1 g put into Bi=1; j ; Bi+1; j;1 ; Bi+2; j ; Bi+3; j;1; : : : ; and Bn; j , for j = 2::n:. - Recall above example as shown in Fig. 3(b), let each element of Go (2) = fS11 ; S13 ; S15 ; : : : ; S49 g, which are allocated into B1;2 ; B2;1 ; B3;2 ; B4;1 ; B5;2 ; B1;3 ; B2;2 ; B3;3 ; B4;2 ; B5;3 ; : : : ; and B5;5 . The final result is shown in Fig. 3(c). describe how to allocate G e = S2n2 g into n n BSs below. The partition of Ge into two subgroups is same with the work in Go , and then we perform the same operation in G1 and G2 steps. An example is shown in Fig. 4(a) and Fig. 4(b). Observe that there is an additional operation is needed to be executed. In the n 2 2-D tori-BSs, each element in each column circularly shifts down one position. This operation is illustrated in Fig. 4(c). An assignment result of G e is shown in Fig. 4(d). As a result, combining with the positioning result of G o and Ge , each BS keeps two distinct data subsegments. An example is given in Fig. 5. We. now. fS2 S4 S6 ;. ;. ;:::;. That is, in our data retrieving operation, these two data subsegments will be used interchangeably as displayed in Fig. 7. In addition, it is easily in Fig. 8 to see that S 1 is located in BS B1;1 on time unit 1, S 2 is in B1;1 on time unit 2, S3 is in B1;1 on time unit 3, etc. Other instance is illustrated in B1;5 ; when S1 is flowed to B1;5 on time unit 2, and therefore S 2 ; S3 ; S4 ; S5 ; :::; S2n2 will be sequentially flowed to B1;5 on the forthcoming time units. Obviously, it is accomplished the data broadcasting operation.. 4 The Proposed VOD Protocol The VOD strategy is based on multi-hop transmission, which is divided into three operations.. . The VOD Operation – For each MH, a VOD operation is proposed to efficiently retrieve video data from BSs. The Error-Handling and Handoff Operations – Two possible cases are discussed for the failure transmission and handoff problems..

(5) 4.1 The VOD Operation The VOD operation herein is equally to build a multicasting tree for multiple MHs since our scheme allows MHs request their VOD demand in the same cell area, and the broadcast protocol guarantees that first data subsegment will be coming to each BS after waiting a period of time. The design difficulty is that each MH’s request time may be different. We now describe the construction of the VOD-multicast tree in same cell area with above consideration. Observe that each cell area can be seen as a MANET. Existing MANET multicast routing protocols [19] [4] [13] [5] [14] [22] [2] can be directly used in our scheme. Existing multicast protocols are classified into proactive/off-line [19][22] and reactive/on-demand [4][13][5][14] approaches. Proactive multicast protocol is to pre-build a shared-tree, and reactive multicast protocol is to construct a tree on-demand. Generally speaking, proactive multicast protocol takes high transmitting time since this approach needs high maintenance cost. The drawback of proactive protocol is not always finding the shortest-path. Therefore, proactive multicast protocol is not very suitable to the topology-changeability network. Two cases of our constructed VOD-multicast tree are discussed.. . . New Call – Existing reactive/on-demand multicast protocols [4][13][5][14] are used for each new call of a MH to connect to a determined BS (the determined rule is explained later), such that the waiting time is minimal. Joining to a Multicast Tree – Existing proactive/offline multicast protocols [19][22] are utilized to connect to a determined BS. A shared-tree is to be pre-builded, which all of the destination nodes of the multicast tree are waiting for first data-segment. Observe that the root of such shared multicast-tree is determined by a determined rule which is explained later. When a MH initiates a VOD request, a join operation is performed to join the shared-tree such that all of the destination node of the original tree and the new member are waiting for the same data-subsegment coming.. In the following, we present the determined rule, which is the rule of determining MH connecting to a BS with minimal waiting time of the first coming data subsegment. Four possible BSs may be connected. For simplicity, let LU ; LD; RU, and RD respectively denote as left-up, leftdown, right-up, and right-down BSs, as shown in Fig. 9. Each of LU ; LD; RU ; and RD should record the current stored two data subsegments, which is denoted as X[i; j], where X 2 fLU ; LD; RU ; RDg; i 2 Go and j 2 Ge : For a BS, suppose that X[i; j] represents the two data subsegments, it is worth mentioned that if let X [i 0 ; j0 ] denote the two new data subsegments which received by performing a row-shift and column-shift operations, so that i 0 = ( j + 1) mod2n 2 and j0 = (i + 1) mod2n 2 : For instance, consider B 1;1 [1; 10] as shown in Fig. 8(a), such that B 1;1 [11; 2] as illustrated in Fig. 8(b) at the next time unit. One important work of the data-broadcasting scheme is to search a feasible path from a MH, which initiates a VOD request, to a BS B, where B 2 fLU ; LD; RU, RDg: The BS B satisfies the following condition; B is with maximum value i or j of X [i; j]; and X 2 fLU ; LD; RU, RDg. Suppose that LU [15; 8]; LD[11; 16]; RU [3; 14], and RD[17; 4] if n = 3 as shown in Fig. 9. We observe that the maximum value is 17, so a MH should flood a path-search packet to search a path to RD.. Figure 9: Four neighboring BSs. Following above description, we assume that the maximum value is M ; while RD keeps value M : We observe that any new MH, which initiates a new VOD request, should connect to RD preserving at least 2n 2 ; M time units: It is because that LU ; LD; and RU will not get larger value than RD for 2n 2 ; M time units. Reminder that each bast station just only keeps the current stored data-subsegment information. Following above example, BSs LU ; LD; RU, and RD just knows [15; 8]; [11; 16]; [3; 14], and [17; 4], respectively. It implies that if any MH communicate to one of their four neighboring bast stations, then this MH just can acquire the information of one of BSs. However, each MH should know the data-subsegment information of all neighboring BSs. We will provide a formula such that if we can know one of data-subsegment information then we can derive three other data-subsegment informations.. . A bit counter is maintained to distinguish the current time unit is odd or even if the bit counter is 0 or 1. Given a pair of neighboring BSs U and D; where U and D located in the same column. Suppose that U [i; j] and D[i0 ; j0 ] denotes the data-subsegment information, two cases are formally stated below. – Case 1: If the bit counter is 0, then i 0 = ( j + 3) mod 2n2 and j 0 = ( j + 1) mod 2n 2 : Inversely, i = ( j0 ; 1) mod 2n 2 and j = (i0 ; 3) mod 2n 2 : - For instance as shown in Fig. 8(a), consider BSs B1;1 [1; 10] and B 2;1 [13; 2], so 13 = 10 + 3 and 2 = 1 + 1:. – Case 2: If the bit counter is 1, then i 0 = ( j + 1) mod 2n2 and j 0 = (i + 3) mod 2n 2 : Inversely, i = ( j0 ; 3) mod 2n 2 and j = (i0 ; 1) mod 2n 2 : - For instance, as shown in Fig. 8(b), consider BSs B1;1 [11; 2] and B 2;1 [3; 14], so 3 = 2 + 1 and 14 = 11 + 3:. . Given a pair of neighboring BSs L and R; where L and R located in the same row. Let L[i; j] and R[i 0 ; j0 ] denotes the data-subsegment information, then i 0 = (i + 2n) mod 2n 2 and j 0 = ( j + 2n) mod 2n 2 : Inversely, i = (i0 ; 2n) mod 2n 2 and j = ( j 0 ; 2n) mod 2n 2:.

(6) Figure 11: (a) The data partition and (b) data transmission of snake-like scheme. Figure 10: Examples of (a) failure transmission, and (b) handoff. - For instance as shown in Fig. 9, if n = 3 and LU [15; 8] and RU [3; 14] are a pair of neighboring base stations located in the same row, 3 = (15 + 2 3) mod 18 = 3 and 14 = (8 + 2 3) mod 18 = 14: For an another instance, if LD[11; 16] and RD[17; 4] are a pair of neighboring base stations located in the same row, 11 = (17 ; 2 3) mod 18 = 11 and 16 = (4 ; 2 3) mod 18 = 14: Based on above formula, we can obtain other three datasegment informations if we just know one data-segment information. If we can acquire the four data-segment information, then we can determine a final path to one of he four BSs in order to wait for the first data-segment incoming. Obviously, the BSs with maximum value of [d 1 ; d2 ] of B[d1 ; d2 ]; where B 2 fLU ; LD; RU, RDg; will be elected to be the connecting BS. This indicates that the waiting time for the first data-segment will be minimum.. 4.2 Error-Handling and Handoff Operations One advantage of our scheme is to provide a fast recovery process if there is failed transmission and the handoff problem. Given a mobile host MH, let MH[d ] represents mobile host keeps Sd data-segment. Two conditions of reconnecting to BSs are stated below. E1: Error-Handling – When the data transmission of a MH is failed, MH may redetermine a new BS such that the new BS has smaller value of d which is most near to d.. Figure 12: The average waiting time of (a) ours and (b) snake-like scheme in a B 55 . This indicates that our scheme and the integrated wireless network model can be used to effectively handle the failed transmission and handoff problems.. 5 Performance Evaluation One possible scheme is using snake-like ordering scheme. A possible of data transmission of snake-like scheme is also outlined as illustrated in Fig. 11(b). A comparison of the data-allocation of our and snake-like schemes is illustrated in Fig. 11(a). A video is partitioned into n 2 data subsegments, each length nl2 ; where l is original video length. Our scheme is split video with length l into 2n 2 data subsegments, such that length of each data subsegment is 2nl 2 : In the following, a data-transferring unit time is assume as U = f ( 2nl 2 ) in our scheme, therefore 2U may be used for the snake-like scheme. The total time Ttotal includes waiting. - For instance as shown in Fig. 10(a), the data transmission of MH[17], or node 10, which connects to RD, is failed, so node 17 will connect to LD since value 16 in LD[11; 16] is near to 17. E2: Handoff – When a MH is leaving out the original cell area and enter into a new cell area, then a handoff process can be easily achieved by let MH connect to a new BS, where BS 2 fLU ; LD; RU, RDg; such that BS has smaller value than d and the proximate value of d : This operation may be connect to a multicast tree if the root of a existing shared-tree is the BS. - For instance as illustrated in Fig. 10(b), a MH[12], or node 10, is roaming to a new cell area, so node 17 will connect to LD since value 11 in LD[11; 16] is near to 12.. Figure 13: The average waiting time of (a) ours and (b) snake-like scheme in a B 77 ..

(7) Figure 14: A comparison table of average waiting time vs. number of BSs.. is varying from 3 to 17. The result of Fig. 14 is obtained by assumed that each cell area exists one MH which wait for the first data-segment S 1 : It is observed that our average waiting time is better than snake-like scheme has. To illustrate the effect of the average waiting time vs. the failure transmission. A simulation is done by making different assumption. Note that Fig. 15 is obtained by assumed that each cell area exists one MH but waits for the data-subsegment S i ; where 1 i 2n 2 : This indicates that the failure transmission is occurred during transmitting the data-subsegment S i : From the result of Fig. 14 and Fig. 15, we observe that the average waiting time of our scheme is better than snake-like scheme. By comparing the simulation result of our scheme from Fig. 14 and Fig.15, it is worth noting that the average waiting time for the failure transmission is approximately equal to the new call has. It implies that the our scheme is a stable strategy. To conclude this section, it is beneficial to adopt our proposed scheme, which is evaluated by our simulation result.. 6 Conclusion. Figure 15: A comparison table of average waiting time vs. number of BSs. time Twaiting and data-transmitting time Ttransmit . Therefore, we have the following result. Ttotal. = Twaiting + Ttransmit :. It is observed that the time cost of Ttotal is mainly determined by value of Twaiting : Therefore, a simulation result is given to make a comparison of Twaiting for our and snakelike schemes. The formula of the average waiting time formula of our scheme and snake-like scheme respectively are 2n2. ∑ ((2n2 ; M + 1) U ). k=1. 2n2. In this paper, we investigate a broadcast-VOD problem in an integrated wireless mobile network. Existing wireless mobile networks can be classified into single-hop (such as GSM system) and multi-hop mobile networks (such as Adhoc network). A single-hop wireless mobile network is characterized by each base station (or BS) connecting to other BS through MSC (mobile station center) by a wired network, but each mobile host (or MH) communicating with its own BS within one-hop transmission radius. Our proposed wireless mobile network is an integration of singlehop and multi-hop mobile networks. The integrated network is characterized by each BS connecting to other BS by a wired network, but each MH communicating with a BS within multi-hop transmission steps. In our integrated network, BSs are connected by a 2-D tori interconnection network. In addition, an important application, called broadcast-VOD protocol, is presented in such an integrated network. Therefore, two contributions are presented in this paper: (a) a new integrated wireless mobile network is proposed, (b) a fast broadcast-VOD protocol is introduced in such a new network. In our strategy, VOD-data are well scheduled and broadcasted among 2-D tori-BSs such that the average waiting time is minimum. In each cell area, a VOD (Video On-Demand) is dynamically requested by any MH at the different time to dynamically construct a multicast tree, while the root node of multicast tree is a BS. This guarantees that each MH can receive the VODsegment from its corresponding BS within minimal waiting time. Performance analysis justified that our proposed scheme outperforms existing schemes.. and n2. ∑ ((n2 ; M + 1) 2U ). k =1. n2 , where M is the maximal value of data subsegment in. fLU LD RU RDg of a cell area. In our simulation, we let ;. ;. ;. unit time U = 100 minute. For example, the average wait2n2 ing time Twaiting of B55 and B77 for our and snake-like schemes are illustrated in Fig. 12 and Fig. 13. A formal comparison table of the average waiting time of our and snake-like schemes are given in Fig. 14, where n. References [1] Charu C. Aggarwal, Joel L. Wolf, and Philip S. Yu. On optimal batching policies for video-on-demand storage servers. In Proceedings of IEEE 1996 Computer Society Press International Conference on Multimedia Computing and Systems, Hiroshima, Japan, pages 253–258, June 1996. [2] Yuh-Shyan Chen, Tzung-Shi Chen, and Ching-Jang Huang. Som: Spiral-fat-tree-based on-demand multicast protocol in a wireless ad-hoc network. In Proc..

(8) of The 15th International Conference on Information Networking (IEEE ICOIN-15), jan 2001. [3] Yuh-Shyan Chen and Kuan-Chang Lai. Mesh: Multieye spiral-hopping protocol in a wireless ad hoc network. In Proc. of IEEE International Conference on Computer Communication and Network (ICCCN 2000), oct 2000. [4] Ching-Chuan Chiang and Mario Gerla. On-demand multicast in mobile wireless networks. In Proceedings of 1998 Sixth International Conference on Network Protocols, pages 262 –270, 1998. [5] Ching-Chuan Chiang, Mario Gerla, and Lixia Zhang. Forwarding group multicast protocol (fgmp) for multihop, mobile wireless networks. Cluster computer, pages 13–16, October 1998. [6] Asit Dan, Dinkar Sitaram, and Perwez Shahabuddin. Dynamic batching policies for an on-demand video server. Multimedia Systems 4(3), 444:112–121, June 1996.. [17] Akiko Nakaniwa, Masaki Ohnishi, Hiroyuki Ebara, and Hiromi Okada. File allocation in distributed multimedia information networks. IEEE 1998 Global Telecommunications Conference. The Bridge to Global Integration, 2:740 –745, 1998. [18] S. Y. Ni, Y. C. Tseng, Y. S. Chen, and J. P. Sheu. The broadcast storm problem in a mobile ad hoc network. In Proc. of Fifth ACM/IEEE International Conference on Mobile Computing and Networking (MobiCom’99), pages 151–162, aug 1999. [19] Elena Pagani and Gian Paolo Rossi. An on-demand shared tree with hybrid state for multicast routing in ad hoc mobile wireless networks. In Proceedings of 1999 International Workshops on Parallel Processing, pages 4–9, 1999. [20] Chunming Qiao, Hongyi Wu, and Ozan Tonguz. icar : an integrated cellular and ad-hoc relay system. In Proceedings of IEEE ICCCN 2000: IEEE International Conference on Computer Communication and Network, Las Vegas, NV, pages 16–18, Oct 2000.. [7] Rohit Dube, Cynthia D. Rais, Kuang-Yeh Wang, and Satish K. Tripathi. Signal stability-based adaptive routing (ssa) for ad hoc mobile networks. IEEE Personal Communications, pages 36–45, feb 1997.. [21] Te-Chou Su and Jia-Shung Wang. On-demand multicast routing sheme and its algorithms. In Proceedings of 1999 13th International and 10th Symposium on Parallel and Distributed Processing, pages 212–217, 1999.. [8] Zheng H., Wang S., and Copeland J.A. Qos aware mobile video communications. In Proceedings of IEEE 1999 Military Communications Conference, MILCOM 1999, pages 1231–1235, 1999.. [22] Y. C Tay, C-K Toh, and C. W Wu. Ad hoc multicast routing protocol utilizing increasing idnumbers(amris) functional specification. Internet Draft, november 1998.. [9] Ren-Hung Hwang and Youn-Chen Sun. Optimal video placement for hierarchical video-on-demand systems. IEEE Transactions on Broadcasting, 444:392–401, Dec 1998.. [23] Meng T.H. Wireless video systems. In Proceedings of 1998 IEEE Computer Society Workshop on System Level Design, pages 28–33, 1998.. [10] David B. Johnson and David A. Maltz. Dynamic source routing in ad hoc wireless networks. Mobile Computing, edited by Tomasz Imielinski and Hank Korth, Chapter 5, pages 81–153, 1996. [11] Sung Bae Jun and Won Suk Lee. Video allocation methods in a multi-level server for large-scale vod services. IEEE Transactions on Consumer Electronics, 444:1309–1318, Nov 1998. [12] Young-Bae Ko and Nitin H. Vaidya. Location-aided routing(lar) in mobile ad hoc networks. MobiCom’98, pages 67–75, aug 1998. [13] Young-Bae Ko and Nitin H. Vaidya. Geocasting in mobile ad hoc networks: Location-based multicast algorithms. Second IEEE Workshop on Mobile Computer Systems and Applications New Orleans, Louisiana, pages 25–36, Feb 1999. [14] Sung-Ju Lee, William Su, and Mario Gerla. Ondemand multicast routing protocol (odmrp) for ad hoc networks. IETF manet(draft-ietf-manet-odmrp02.txt), 2000. [15] Elizabeth M.Royer and Chai-Keong Toh. A review of current routing protocol for ad hoc mobile wireless networks. IEEE Personal Communication, pages 46– 55, April 1999. [16] Davies N., Finney L., Friday A., and Scott A. Supporting adaptive video applications in mobile environments. IEEE Communications Magazine, 366:138– 143, June 1998.. [24] Chai-Keong Toh. Associativity-based routing for ad-hoc mobile networks. Wireless Pers. Commun., 4(2):1–36, Mar 1996. [25] Y. C. Tseng. A dilated-diagonal-based scheme for broadcast in a wormhole-routed 2d torus. IEEE Transactions on Computers, 46(8):947–952, August 1997. [26] Dongyan Xu, Baochun Li, and Klara Nahrstedt. Qosdirected error control of video multicast in wireless networks. In Proceedings of 1999 Eight International Conference on Computer Communications and Networks, pages 257–262, 1999..

(9)