A Time and Distance - Based Multicast Algorithm for IPv6 Mobile Networks

Transcription

1 A Time and Distance - Based Multicast Algorithm for IPv6 Mobile Networks Wu-Hsiao Hsu, Bin-Hau Lee, Ming-Han Liu, Bor-Yeh Shen, Kun-Hua Lin Department of Computer Science and Engineering, Ming-Chuan University, Taoyuan, Taiwan wuhsiao@mcu.edu.tw Abstract IP multicasting has naturally been considered the ideal technique to be used with multimedia communications. Unfortunately, current multicast protocols do not consider the dynamic membership. Therefore, we propose a time and distance-based multicast algorithm TDBMA for IPv6 mobile networks. The TDBMA is subject to time and distance. A foreign network is qualification to join the existing multicast tree only its value of time and distance is larger than a computed threshold. The experimental analyses are presented to characterize the performance of the proposed algorithm. Keywords: IP multicast, Mobile networks, multicast tree. 1. Introduction Recently mobile hosts MH and wireless networks have become widely popular. For the wireless network, users are allowed to connect networks without being tethered off a wired network. In the meantime, multimedia applications such as video on demand VoD, video conferencing, and on-line games have also become extremely popular and attractive. It is becoming clear that MH will expect to have access to the multimedia applications available in traditional wired networks. Therefore, multicasting has naturally been considered the ideal technique to be used with multimedia communications, mainly because its inherent nature is to minimize the network resource. Consequently, many efforts are being made to provide efficient mobility and multicasting support in the current IP networks. Many multicast protocols, such as DVMRP, PIM-SIM, and MOSPF, have been proposed by the IETF and used on the Internet. These protocols implicitly assume static hosts when building a multicast delivery tree. They do not consider the dynamic membership. Therefore, current multicast protocols have not been optimized to handle host mobility. The addition of mobility in the host group implies that the multicast algorithm ought to focus not only on the issue of dynamic group membership, but also on host location. When a MH moves to a foreign network and resubscribes to a multicast tree, the foreign network must determine whether the subscription was accepted or rejected. If the foreign network accepts the subscription, it is clear that frequent reconfiguration of multicast delivery tree every time a host moves could result in substantial control overhead. On the other hand, if the foreign network rejects the subscription, it is also clear that leaving a multicast tree unchanged every time a host moves can result in inefficient and increased delay of multicast datagram delivery. In this paper, we propose a time and distance-based multicast algorithm TDBMA for IPv6 mobile networks. When a MH arrives at a foreign network and resubscribes to a multicast group, the foreign network can use TDBMA to determine whether it can join the multicast group or not. TDBMA is subject to two key factors: one is time which describes the total time that the MH stays in the foreign network. The other is distance which describes the distance from the foreign network to the multicast tree. The foreign network can join the multicast group directly only both the time and distance are larger than a computed threshold. The remainder of this paper is organized as follows. In section 2, the previous works was introduced. Section 3 introduces our TDBMA. The simulation model and results are described in section 4. Some conclusions are given in section Previous works Previous works on finding a multicast tree can be classified into two categories. First, the destination nodes are static during the transmission period. This multicast routing problem has been formulated as finding a minimum cost multicast tree problem, which is also known as the Steiner tree problem [1], [2] and was shown to be NP-complete [3]. Secondly, a dynamic network node denoted as MH in this paper is allowed to move during its call holding time. Two basic mechanisms have been proposed by the IETF in mobile IP to support multicasting. These are known as 1 Proceedings of the 18th International Conference on Advanced Information Networking and Application AINA 04

2 remote subscription and bi-directional tunneling, and have been described in [4] for IPv4 and [5] for IPv6. The remote subscription provides the shortest routes for delivery of multicast datagrams to mobile hosts, while the bi-directional tunneling hides host mobility from all other members of the group. The bi-directional tunneling is based on home agent functionality. Mobile hosts receive multicast data by way of their home agents. This results from the fact that multiple Mobile IP tunnels from different HAs terminate at a common FA. This feature is called the tunnel convergence problem. MoM [6] was proposed to solve this problem. It uses the DMSP Designated Multicast Service Provider to avoid duplicate datagram being tunneled to the common FA. RBMoM [7] was proposed and intended to trade off between the shortest delivery path and the frequency of the multicast tree reconfiguration by controlling the service range of the multicast home agent. 3. The time and distance - based multicast algorithm TDBMA When a MH moves to a foreign network FN, TDBMA must be used by the foreign network router FNR to determine whether it can join the multicast tree or not. For simplicity, we assume that an existing multicast routing protocol DVMRP, PIM-SIM, or MOSPF was used in the wired network and already established a multicast tree T. The source node and destination nodes of the multicast group mg are described by a pair, D where is the source node of the multicast group mg and D is the finite set of destination nodes to be connected from source node. T = V T, E T where V T is the set of nodes on the tree, {} D V T and E T is the set of links on the tree. 3.1 The description and definition of the time and distance When a MH arrives at a FN, it follows the registration scenario of mobile IPv6 that sends a packet via FNR to its HA containing a Binding Update BU destination option. If the MH resubscribes to the mg, TDBMA must immediately be used by the FNR to determine whether the FNR is qualification to join the T or the MH just receive multicast packets by way of its HA, using unicast mobile IP tunnels from its HA. The key points that determine to join the T or not depend on time and distance. The time is defined by how long the MH stays in the FN. It is possible that the MH may temporarily stay in the FN and move to another FN as soon as possible. In this case, the FNR is not qualification to join the T. On the other hand, the MH may stay in the FN for a period of time. During this period of time, if the MH wants the services of the HA beyond the current registration period, the MH must send a new BU to it before the expiration of this period. As a result, the more the BU was sent by the MH, the longer the MH stays in the FN. In this case, it is inefficient for the MH to receive multicast packets by way of its HA. The FNR may join the T and then directly forward the multicast packets to MH. In contrast to the time, the distance is defined by the shortest path length from MH to the T. If the distance from the MH to the HA is larger than the distance from MH to the T, it is also inefficient for the MH to receive multicast packets by way of its HA. The best way for the MH to receive multicast packets is directly from the T. In this case, the FNR may join the T and forward the multicast packets to the MH. 3.2 The description of the TDBMA In TDBMA, two tables have been defined. As shown in figure 1, each FNR not including in the T maintains a Binding Table BT which describes the information of a MH, the obtained care-of address CoA of a MH, the total number of Binding Update BU that a MH has sent denoted by BUG, and the multicast group MG that a MH is going to join. Also, the of each T records a Network Table NT which describes the FNR, BUG, the average number of BUG for a multicast group denoted by ABUG, and the multicast group MG that the FNR is going to join. TDBMA uses a FN-metric, referred to as foreign network metric, to determine whether a FNR is qualification to join the T or not. FN-metric can be defined as where d t FN-metric = d + 1- t 1 if D _ FH D _ FS 0 if D _ FH D _ FS, and if BUG ABUG 1 0 if BUG ABUG. Both d and t represent the distance and time, respectively. D_FH and D_FS denote the distance from the FNR to HA and the distance from the FNR to the ontree nodes, respectively. For the distance d, if D_FH is larger than D_FS, this means that the MH may be far away from its HA and close to the T. In this case, the value of d is set to 1. Similarly, for the time t, if BUG is larger than ABUG, this means that the MH may stay in the FN for a long time. In this case, the value of t is set to 1. TDBMA recommend that the FNR can directly join the T only both the value of t and d is equal to 1. is a constant weighting factor, where 0 1. Choosing a value for close to 1 and 0 makes the FN- 2 Proceedings of the 18th International Conference on Advanced Information Networking and Application AINA 04

3 metric to d and t, respectively. For other cases, the FNmetric depends on the value of d and t. 3.3 The operation of TDBMA As shown in figure 2, suppose the, D = A, {C, D, E, F} and both MH1 and MH2 have moved from nodes F to H and D to G, respectively. For the initial stage, each MH will receive multicast packets by way of its HA after finishing the registration procedures. Suppose MH1 has sent totally six times of BU to its HA F and MH2 has sent totally two times of BU to its HA D. In the meantime, each FNR will send its own BT periodically to the source node A. For example, Both H and G will send the BT to the A. After receiving all BTs, the will calculate ABUG and write the result into NT for the multicast group mg. Let T BUG = {BUG 1, BUG 2,.., BUG x } denote as the set of BUGs of all nodes not in the T. Also, letµ BUG denote as the mean of the T BUG containing X items and can be represented as x i BUG 1 BUGi where BUG i denotes the sum of the values of all data in the T BUG. The corresponding standard deviation BUG can also be defined as BUG X x BUGi i 1 X The proportion of data falling within k standard deviations of the mean, which is called Chebyshev s theorem [8], is given by 1-1/k 2 where k > 1. The important advantage of Chebyshev s theorem is that the desired proportion is independent of the distribution of data set. From Chebyshev s, an ABUG can be calculated as following, BUG ABUG = µ BUG + k BUG. For example, as shown in figure 2, the T BUG = {6, 2}, µ BUG = 4 and BUG = If we let k = 1.1, ABUG = Obviously, the value of ABUG depends on a given k. For calculating both D_FH and D_FS, it is easy to find the D_FH by checking the unicast routing table resided in the FNR. The unicast path is typically the shortest path in terms of hop length. For example, as shown in figures 3a and 3b,, D = A, {D, E, F, H} and V T = {A, B, C, D, E, F, G, H}. The HA and FNR of the MH1 are nodes H and K, respectively. When the 2 MH1 arrived at the node K, node K will find the D_FH, which is 2 hops and the path is KLH, by checking its unicast routing table directly. In contrast to D_FH, D_FS is found by sending a probe message from FNR to the T along the unicast shortest path from FNR to the. If an on-tree node receives this probe message, it will reply to this message which includes the information of D_FS. Otherwise, the probe message is again sent towards the. For example, as shown in figure 3c, node K will find the shortest path to by check its unicast routing table. Then it will send the probe message towards the. The on-tree node G, as shown in figure 3d will reply that the D_FS is 1 to node K. After determining the d and t, the FN-metric was calculated by the FNR as following, FN metric 1 0 if d 1 and t 1 and 0 and 1 otherwise. Obviously, if FN-metric is equal to 1, the FNR is qualification to join the mg. That is, the FNR can join the T and directly receive all multicast packets. Otherwise, the FNR is not allowed to join the T and will still receive the multicast packets by way of its HA. For example, as shown in figure 2, only node H is qualification to join the T since BUG > ABUG = 6 > and D_FS 1 < D_FH 2. If we let k = 1.5, none of the FNRs are allowed to join the T. It is clear that if a smaller k is given, there are more nodes to be considered to join the T. 3.4 The TDBMA In TDBMA, when a MH moves to a FN and resubscirbes to a mg, there are four cases to be discussed : 1 Both the HA and FNR have already joined the mg. In this case, after the procedure of registration, the MH can receive the multicast packets by way of the FNR. As shown in figure 4a, suppose the, D = A, {D, F, H} and MH1 moves from nodes F to H. Since both HA F and FNR H of the MH1 have joined the mg, MH1 can receive the multicast packets by way of the H. 2 The HA has joined the mg but the FNR does not. In this case, the FNR must calculate its FN-metric. As shown in figure 4b, suppose the, D = A, {D, F}. Sine node F has joined the mg but node H does not, node H must calculate its own FN-metric in order to determine whether it can join directly the mg or not. 3 The FNR has joined the mg but the HA does not. In this case, the HA must be forced to join the mg because it is possible that the MH will move to another FN at any time. At this time, the MH can still receive the 3 Proceedings of the 18th International Conference on Advanced Information Networking and Application AINA 04

4 multicast packets by way of its HA. As shown in figure 4c, suppose the, D = A, {D, H}. Because the node F does not join the mg, node H must inform the node F to join the mg. 4 Both the HA and the FNR does not join the mg. In this case, the HA must be forced to join the mg and the FNR can calculate its own FN-metric in order to determine whether it is qualification to join the mg or not. As shown in figure 4d, suppose the, D = A, {D}. The nodes F and H are forced to join the mg and calculate its own FN-metric, respectively. 4. The simulation model and results 4.1 Simulation models and assumptions A network model with n nodes, which is similar to that of [9], is randomly distributed in a rectangular coordinate grid. Each node is located in an integer coordinates and represented as a multicast router and base station. The radius of transmission range of each node is chosen randomly from 0, L]. The maximal node n and radius L are set to 50 and 100, respectively. The size of the number of moving hosts will be given by 0 initially and increased by 5 each time until 20. Also, the initial size of multicast member D is given by 20. For simplicity, suppose the distance is measured by hop count. That is, the distance of a path from MH to source node is measured by the number of routers that a datagram encounters along this path. The BUG of a BT maintained by the FNR can be chosen randomly from 1, 30. The k is set to 1.1 for our all experiments. The is given by 0, 1, and 0.2 for different performance evaluation. All parameters are summarized in the table Simulation results and discussion The simulation results are shown in figures 5 and 6 which are to evaluate the number of nodes on the multicast tree T and the number of operation of joining and pruning, respectively. Three different algorithms, Join remote subscription, Tunnel bi-directional tunneling, and TDBMA are compared by changing the parameter to 0, 1, and 0.2 respectively. The total number of nodes in the V T is 27 based on the given D for our simulation. For = 0, FN-metric = t. That is, the value of FNmetric is determined by t factor only. Look at figures 5a and 6a, if the total number of moving node is 0, TDBMA produces the highest nodes in the V T because the MHs, which possible already arrived at FN, are randomly chosen. If the HAs of these chosen MHs do not join the T, these HAs are forced to join the T. Hence even there are no MHs, the nodes in the V T will still be increased. If the total number of moving nodes is increased from 5 to 20, TDBMA still almost produces the highest nodes in V T. This is because there are too much pruning operations in the Join algorithm. For example, for the number of moving nodes 5, the number of join operation is 3. The remaining 2 moving nodes are moved to the on-tree nodes. Hence, the total number of nodes in the V T is 30. But since the pruning operation is also 3, the number of nodes in the V T is reduced to 27, which is same number as our simulation default. On the other hand, since the Tunnel algorithm is independent of the moving host location, the number of nodes in the V T is never changed and the number of joining and pruning is always 0. For = 1, FN-metric = d. That is, the value of FNmetric is determined by d factor only. Look at figures 5b and 6b, since there are higher number of joining and pruning than that of = 0, the Join algorithm has the lowest V T. TDBMA still has the highest nodes in the V T. The reason is the same as the = 0. For = 0.2, FN-metric depends on t and d. Look at figures 5c and 6c, it is obviously that the number of nodes in the V T is reduced for TDBMA because it is not easy for each MH to satisfy the condition of FN-metric. From our simulation, the Join algorithm produces too much number of Joining and Pruning in high mobility. As a result, the multicast tree T reconstructs more frequently. Hence, the multicast network established by Join algorithm is quite unstable and also waste more network resource than other algorithms. On the other hand, although TDBMA produces more nodes in the V T than Join algorithm, the total number of Joining and Pruning is lower than that of Join algorithm. Hence, the multicast network is more stable and the transmission of multicast packets is more efficient. Finally, though the Tunnel algorithm has the most stable multicast network, it has the lowest efficiency for transmitting the multicast packets. 5. Conclusions When a MH moves to a FN and resubscribe to the mg, an algorithm must be used to determine whether this MH can join the current multicast tree T or receive the multicast packets by way of its HA. In this paper, we propose a new algorithm, TDBMA, which depends on the time and distance of the MH. Unlike Join algorithm, which joins the T without any restriction, and Tunnel algorithm, which the number of tunnels is unlimited, TDBMA can be used by the MH to join the T only the time and distance of the MH satisfy the defined condition. The simulation results show that the multicast tree update by TDBMA is not as frequent as Join algorithm. Also, the TDBMA is more efficient than Tunnel algorithm for receiving the multicast packets. References 4 Proceedings of the 18th International Conference on Advanced Information Networking and Application AINA 04

5 [1] R.Garey and S. Johnson, Computers and Intractability: A Guide to the Theory of NP-Completeness, Freeman, New York, [2] P. Winter, Steiner problem in networks: A survey, Networks, vol. 17, pp , [3] R.Garey and S. Johnson, The rectilinear Steiner tree problems is NP-complete, SIAM J. Appl. Math., vol. 32, pp , [4] G. Xylomenos, and G. Polyzos, IP Multicast for Mobile Hosts, IEEE Commun. Mag., vol. 35, no. 1, Jan. 1997, pp T/FUXPSL 5BCMF / #6 #6. #5 H H. #5 # [5] C. Bettstetter, A. Riedl, and G. GeBler, Interoperation of Mobile Ipv6 and Protocol Independent Multicast Dense Mode, Proc. Wksp. WL Net. Mob. Comp., Toronto, Canada, Aug [6] T.G. Harrison, C.L. Williamson, W.L. Mackrell and R.B. Bunt, Mobile multicastmom protocol: Multicast support for mobile hosts, in: proceedings of ACM/IEEE MOBICOM 97. T#JOEJOH 5BCMF. $P #6.. *1 H #6. T#JOEJOH 5BCMF. $P #6.. *1 H Figure 2. The operation of TDBMA [7] Chunhung Richard Lin, Kai-Min Wang, Scalable Multicast Protocol in IP-Based Mobile Networks, Wireless Networks 8, 27-36, probe messge find the shortest path by check unicast routing table [8] D.R. Byrkit, Statistics Today: A comprehensive Introduction, Menlo Park, California, [9] B.M. Waxman, Routing of multipoint connections, IEEE J. Select. Areas Commun., vol.6, no.9, pp , Dec $P #6. * +, - a MH1 moves from node H to node K. * +, - b Node K calcuates the D_FH by checking its unicast routing table B 5IF#JOEJOH 5BCMF * +., - c Node K calcuates the D_FS by sent the probe message along the unicast shortest path from node K to source A * +., - d The on-tree node G reply to the probe message to node K which include the information of D_FS / #6 #6. Figure 3. The calculation of D_FS C 5IF/FUXPSL 5BCMF Figure 1. The defined tables. B CPUI BOE IBWFKPJOFE UIFTBN FN VMUJDBTUHSPVQ. C IBTKPJOFEUIFNVMUJDBTU USFFCVU EPFTOPU DBMDVMBUFJUTPXO / N FUSJD. D IBTKPJOFEUIFNVMUJDBTU USFFCVU EPFTOPU JTGPSDFE UP KPJO UIFN VMUJDBTUUSFF. JTGPSDFE UP KPJO UIFN VMUJDBTUUSFF DBMDVMBUFJUTPXO / N FUSJD E #PUI BOE EPFTOPUKPJOUIF N VMUJDBTUUSFF Figure 4. The four cases of TDBMA 5 Proceedings of the 18th International Conference on Advanced Information Networking and Application AINA 04

6 Table 1. The simulation parameters Paramete Description Value r n The total number of nodes in our 50 network model L The radius of transmission range 100 between two nearby nodes. MH The total number of moving hosts 0 ~ 20 d The distance of a path from source Hop count to destination t The time that MH stays in a FN 1 ~ 30 k a constant weighting factor 1.1 a constant weighting factor 0, 1, and 0.2 D The size of multicast member 20 /VN CFSPG/PEFTJOUIF. VMUJ U5SFF D B T 5P BM/ U V NSPG+PJOJOHBOE1 C F SVO O J 5PU M B VNCFS / PG+PJ JOHBOE1SVOJOH O /VNC SPG.PW F OH/P J ETF a = 0 /VNCFS PG. PWJOH /PEFT b = 1 /VNCFSPG.PWJOH/PEF T /VNCFS PG/PEFTJ OU F I.VMU BT J UD SF 5F a = 0 5PUBM/ VN CFS PG +PJOJOHBOE1SVO OH J /VNCFS PG. PWJOH /PEFT /VNCFS PG. PWJOH /PEF T b = 1 c = 0.2 Figure 6. The total number of joining and pruning /VNCF S PG/PEFTJ OU F I.VMU JTUD SFF B5 /VNCF S PG. PWJOH /PEFT c = 0.2 Figure 5. the number of nodes in the multicast tree 6 Proceedings of the 18th International Conference on Advanced Information Networking and Application AINA 04