An efficient dynamic multicast agent approach for mobile IPv6 multicast

 

Hong-Ke Zhang*, Bing-Yi Zhang**, Bo Shen

College of Electronics and Information Engineering, Beijing Jiaotong University, Beijing 100044, China

*hkzhang@center.njtu.edu.cn ,**bingyizhang@163.com

 

Abstract

 

 For providing multicast serve in mobility IPv6 network, two approachs have been proposed by IETF, i.e.MIP-BT (mobile IP-bi-direction tunnel) and MIP-RS (mobile IP-remote subscription). But there are some disadvange about them. The approach of MIP-BT has the problem of triangle route and tunnels congregation. The approach of MIP-RS has the problem of much handover latency and handover cost. The other present approachs are also cannot solve above problems effectively. So a novel approach named DMA (dyniamic multicast agent) is proposed in this paper. This approach integrats the merit of MIP-BT and MIP-RS and conquer their disadvange. It can reduce the times of multicast tree restructuring and allow MNs optimize multicase routes simultaneously through selectiong a new multicast agent dynamically. In addition, in this approach MN is left out of the most management and the load in MN is low. MN in the same subnet may select different DMA and the load in each DMA is lightened.The details is described in this paper including the throughness flow chart, the operation of MSA (multicast subnet agent), dynamic multicast agent selection and etc.. The simulation in NS is completed as well. The results in simulation denote that DMA approach can effectively avoid frequent modification multicast tree and the multicast packages can be forwarded through sub-optimal path. The DMA approach can improve the network performance effectively and adapt to the mobile IPv6 multicast.

Keywords: mobile IPv6, multicast, mobile IPv6 handoff, dynamic multicast agent, multicast subnet agent

1 Introduction

  IPv6 is a very significant topic for NGI (next generation Internet). Along with the development of the wireless IP technologies, more and more people have the chance to use wireless equipment accessing network. They will like this method gradually and being in the habit of accessing wireless network. Now there are more than 200 millions of mobile terminal users such as mobile phone users in China. Mobile IPv6 will act as an important role in NGI in future. Multicast is an efficient way for forwarding data from one node or multi-nodes to multi-nodes.The technology of multicast can improve the transmission efficiency and reduce bandwidth cost dramatically. Multicast also becomes a hot point and many research have been studied in it. There have been many multicast route protocols, such as DVMRP (distance vector multicast routing protocol) [1], MOSPF (multicast open shortest path first) [2], MSDP (multicast source discovery protocol) [3], BGMP (border gateway multicast protocol) [4], PIM-DM (protocol independent multicast-dense mode) [5] and SSM (source specific multicast) [6] and so on. Unfortunately they are not suit to mobility wireless network. Supporting the mobility and multicast will becomes the necessary function in NGI. Therefore how to support multicast in mobile IPv6 network is an inevitable issue. The research in mobile multicast is important and significative.

For providing multicast serves to host in IPv4 network environment, some multicast approaches have been proposed such as MoM (mobile multicast protocol) [7], MMA (multicast by multicast agent) [8] and MA (multicast agent)[9]. MoM is based on MIP-BT and try to conquer tunnels congregation through selecting a DMSP (designated multicast service provider) from a set of HA. Howevery, it still need to set up a tunnel between DMSP with FA and cannot solve the triangle route yet. In addition, it need to selectDMSP frequently when MN moving fast. Due to the frequent selection of DMSP, the cost will increase and data package may be lost.

MMA includes MA(multicast agent) and MF(multicast forwarder). Where the function of MA is to join multicast group. Then MN select a MA from a set of MA as a MF and accept multicast datagram from the MF. So This approach can reduce data delivery path length and decreases the amount of duplicate copies of multicast datagrams through selecting a MF and making the route optimal. However it still exists the problem of triangle route.

In MA approach the multicast agent act as the access point of multiple foreign networks to multicast backbone. The multicast agent joins the multicast groups and tunnels multicast packets to the FA. For each multicast agent serves multiple foreign networks, It avoids the frequent modifications of multicast trees.

The above approaches are all proposed in IPv4 network environment and not suit to IPv6 network environment. They all cannot avoid the problem of triangle route.

   The mobility support for IPv6 protocol [10] specifies two basic methods for mobile multicast: (1)  a bi-directional tunnel from a MN to its HA (home agent), called MIP-BT (mobile IP-bi-directional tunnel) and also called HS(home subscription); (2) a (local) multicast router on the foreign link being visited, called MIP-RS (Mobile IP-Remote Subscription). In MIP-BT, the MN tunnels its multicast group membership control packets to its HA, and the HA forwards multicast packets down the tunnel to the MN [10]. In MIP-RS, the MN must use its care-of address and must not use the Home Address destination option when sending MLD (multicast listener discovery) packets [10,11].The two basic methods can retain multicast communications when MNs move. However some issues still exist.

(1) MIP-BT suffers the triangle route which is composed of MN-HA tunnel and HA-S multicast tree path. When the MN is far from its HA, the data forwarding path of multicast becomes        deteriorative.

(2) Multiple tunnels from a subnet to a HA are established in MIP-BT when some MNs that come from the same home link attach at one AR (Access Router) in the subnet and these MNs join the same multicast group at the same time. This case is called tunnels congregation which leads to more network resources being consumed.

(3)  Although the multicast path in MIP-RS is optimal, the frequent handoffs of a MN, which are due to the movement of the MN among subnets, produce much latency. This is because the     handoff action makes the MN leave and re-join the multicast group.

It is interseting the two basic methods for supporting mobile multicast have complementary advanges, so the improved efficiency can be achieved by combining the two approaches. For there is lack of research to conquer the above issues for mobile multicast in IPv6 environment, we propose an efficient approach named DMA (Dynamic Multicast Agent) which accepts the merits of MIP-BT and MIP-RS. DMA uses movement based method and distance based method to select a new multicast agent dynamically, and the new selected agent is responsible for forwarding multicast data to the MN. Such a configuration allows MNs optimize multicast routes and meanwhile reduce the number of switches. In addition to weakening the triangle route problem and diminishing the influence of switch to multicast, this approach provides global mobility in Internet with no restriction on network topologies. In this approach MSA (multicast subnet agent) and DMA attend multicast mobility management. MN left out of the most management and load in MN is low. MN in the same subnet may select different DMA, so load in each DMA is lightened.

In the following sections, we first introduce the overview and description of the DMA. Then, we describe the detail of operations of a MSA (Multicast Subnet Agent), DMA selection process and algorithmic flow chart. At last we complete the simulation and show our approach being efficient.

2. Dynamic Multicast Agent for Mobile IPv6 Multicast

2.1 DMA overview

In this paper, three concepts are defined: MSA (multicast subnet Agent), DMA and Tunnel State.   MSA is the only multicast access point in one subnet. It is responsible for forwarding multicast datagrams to every MN that visits its subnet, and the Access Router running multicast protocol   in a subnet can act as a MSA. MSA is in charge of local dynamic group membership management information via MLD protocol, and periodically sends query messages to solicit reports of all   multicast addresses that are being listened to by local hosts. Tunnel State is a flag in every entry of multicast route table that is maintained by MSA. It represents the state of each group G (record   -s whether the multicast packets are received through tunnel). The simple DMA description is as follows:

(1)       When a MN first joins a multicast group through a MSA, MSA becomes the current DMA

of the MN. Before a new DMA is selected, the current DMA receives multicast data from the tree.

(2)       When the MN leaves the subnet which its DMA is in and attaches at a new subnet, the MSA   

in the new subnet gets multicast data through the tunnel from the MSA to the MN's DMA, and then forwards the data to the MN. The current MSA exchanges MLD information with the MN's DMA through the tunnel also. Each MN corresponds with a DMA which is selected dynamically according to passed path of MN. The process is described in figure 1.

 

Figure 1 Operation of DMA approach

The DMA selecting method and conditions are described in the next sub-section. A MSA will join multicast tree directly if it is selected as a DMA. Dynamic selection of DMA can optimize multicast transmit paths due to the shortest path from DMA to multicast source. Excessively long tunnels can be avoided because of the short distance between DMA and subnet. The DMA can make good use of network resources and avoid choke points of the network, compared with static subnet agent. Tunnel State makes sure that there is only one multicast agent providing service, which corresponds with MNs of a new-join multicast group in a subnet. It can also avoid the tunnel congregation problem. When a MN moves from a subnet to another subnet, it does not need to rejoin multicast trees due to DMA. So the impact on the multicast tree is reduced. In addition, MSA and DMA complete the most work of multicast tree restructuring and the load of MN is low. For each MN has a individual item and different MN can select different DMA, the load of one DMA can be lightened. The detail will be given in following section.

2.2 Operation of MSA

Each MSA maintains a Visitor Table shwn as figure 2. There are two elements kept in every entry of the Visitor Table: MN item and DMA item. The meanings are:

(1) MN item records the mobile nodes that visit the subnet the MSA belongs to and need to be served;

(2) DMA item records which MSA is selected as the MN's DMA.

 

 

 

Figure2. Visitor table

On arriving at a new foreign network, a MN obtains a new CoA (care-of address). The MN registers its current CoA with its HA, and then the MN immediately sends message to its current subnet's MSA. The message mainly contains the MLD group membership report message and   the IP address of pMSA (previous subnet's MSA). When receiving the MLD group membership report message that is sent by a new visitor for group G, the MSA should give some judges including whether there has being the information of multicast group, whether there has being the MSA visitor table, whether there has being the MN item record and so on, and the detail operations of the MSA are as follows:

(a)     If the MSA already has an entry for group G in its multicast route table , add MN to the entry's  

outgoing interface list.

(a.1)examine the Tunnel State. If the Tunnel State is 'YES', it indicates that the MSA has already  

created the tunnel for the group and is receiving multicast datagrams via the tunnel. In this case, it only requires forwarding the MLD group membership report message to the other end of the tunnel and examine the entry related to MN. If there is not tunnel it examine the entry relate to MN directly.

(a.1.a) If the entry related to the MN exists in the MSA's Visitor Table, then the MSA is holding it.

(a.1.b) Otherwise if there is no entry related to the MN in its Visitor Table, it creates a new entry for

the MN. In order to optimize the delivery path, the MSA of the current subnet is selected as the DMA of a MN according to certain rules. In this case, current subnet's MSA adds an entry in its MSA list (detailed in next sub-section) to record the movement path of the MN. And then the current subnet's MSA communicates with the pMSA to obtain the IP address of the previous DMA, and informs the previous DMA to delete all data structures that are related to the MN.

(b)     If the MSA has no entry for group G in its multicast route table, it indicates that the MN is the  

first group member of group G in the subnet, then MSA creates a new entry for group G in its multicast route table and adds the MN to the entry's outgoing interface list.

(b.a) If there already exists an entry related to the MN in the MSA's Visitor Table (other multicast

group), only when the MSA of current subnet acts as the DMA of the MN in this entry, the MSA simply sends join messages to the multicast tree. Otherwise if the DMA is not the MSA of current subnet, then set the Tunnel State to 'YES', create tunnel with the DMA of MN, and forward the MLD group membership report message to the other end of the tunnel.

(b.b) If there is no entry related to the MN in its Visitor Table, the MSA creates one entry for the MN, communicates with the pMSA to obtain the IP address of the previous DMA, and then communicates with the pDMA. If the pDMA doesn't re-select a new DMA, the MSA       adds the pDMA to the entry as the MN's current DMA, sets the Tunnel State to 'YES', creates tunnel with the current DMA of MN, and then forwards the MLD group membership report message to the other end of the tunnel. If the pDMA re-selects a new DMA,        the MSA of current subnet works as the new DMA, the MSA simply sends join messages to the multicast tree, and at the same time, the current subnet's MSA creates and maintains an entry for the MN in its MSA list.

  For describing the operation of MSA more clearly, a throughness flow chart will be shown in figure 4 in following section.

  A MN group member doesn't send a leaving message when it departs from the current subnet to visit another; the MSA detects its departure by the timeout of timer. When the MSA detects that a MN is departing from the current subnet, it deletes the entry maintained for the MN in its Visitor Table. For each multicast group G which the leaving MN has joined, the MSA deletes the MN from the group G entry's outgoing interface list.

2.3 Dynamic Multicast Agent Selection

In the approach, the way of DMA re-selecting is one of key points. The following selection principles are based on both movement and distance [12][13], and the selection of DMA can be performed by a MN's current DMA or MSA. The following description illustrates the dynamic multicast agent selection process by DMA. Each MN corresponds to only one DMA at one time. DMA is a multicast router which participates in Internet routing, provides multicast service for MN, and acts as an access point for the MN to the Mbone. Each multicast router which works as a DMA for an MN maintains a table and computes DMA re-selection for an MN which selects it as its DMA according to a specific rule. The table changes dynamically with the route which the MN passes through. The corresponding DMA, therefore,is selected dynamically with the position of the MN. That is why it is called dynamic multicast agent (DMA). The MSA list which the corresponding DMA maintains the following records:

 

Figure 3 DMA maintenance

   The meaning of the element in the MSA list is as follows: MN item records the mobile node which selected the MSA as DMA; MSA item records MSA of each subnet which MN passes through; Increment item records maintain the path increment of the MSA.

   The elements of forming MSA list can be divided into three categories:

(1)       If the subnet is the first subnet which the MN passes through, the MSA of this subnet works   

as DMA, and is the initial DMA of the MN. In this case, it creates and maintains an entry for this MN in the MSA list of this MSA.

(2)       When a MN enters a new foreign subnet, the MSA of this subnet receives MLD group

member report messages of the MN. If the MSA has a group member of this multicast group, it receives multicast packets and forwards them to the MN. If there is no entry about this MN in the MSA Visitor Table, it selects the MSA of current subnet as DMA of MN to optimize the traffic route. In this case, it creates and maintains table for this MN in the MSA list of this MSA, and announces the previous DMA to delete all data structures of related MNs.

(3)  When a MN joins a new foreign subnet, MSA receives MLD group member report message of the MN. If the MN is the first group member in the subnet and there is no element of this MN in the MSA Visitor Table, the MSA communicates with the previous DMA of MN. If the previous DMA selects a new DMA, the MSA of current subnet works as new DMA, and it creates and maintains an entry for this MN in the MSA list of this MSA.

  When the entry is created in MSA list for a MN, the MN and its current subnet MSA are written to the location of DMA. Whenever the MN enters a new foreign subnet, the new subnet MSA will communicate with the DMA of the MN. If the DMA receives new messages from the new foreign subnet to inform that a new DMA selection has been taken place, the DMA deletes the entry maintained for the MN; otherwise the DMA will look up the entry list and determine if there is a    record for it. If it is null, the DMA adds the new foreign MSA to the MSA list. When the summation of the path increment of MSAs in MSA list reaches the assigned threshold, DMA re-selection   occurs. At this point, the old DMA sends a message to the MN, and then the MN gives notice to the MSA of the subnet it belongs to. Once the current MSA receives the message, it will choose itself as the new DMA of the mobile node, and add an entry for the MN in its MSA list. The path increment of a MSA is defined as follows:

                      [Length(MN-DMA)/Length(DMA-S)]                           (1)

 where [x] is the minimum integer greater than or equal to x, and Length(y) is the distance of the path y. If the MSA needs to receive multicast packets via the DMA, it creates a tunnel between the MSA and the DMA to establish the existence of the membership. If the DMA has been the member of the group, the DMA will add a link to the MSA in the output interface list. If not, it need to request to add the group. When the DMA receives the multicast packets, it will forward them to all the members of the group through local link, and forward them to other MSA through a tunnel. The whole flow chart can be seen in figure 4.

 

 

Figure 4 Flow chart of MSA

 For giving a clear picture of how the switch of DMA occuring, we describe the switch as follows. On arriving at a new subnet, MH will sends a message to its MSA in current subnet. The message includes the IP address of previous MSA. Then MSA in current subnet will communicate with the previous MSA. From the Visitor Table in previous MSA, current MSA can obtain the IP of the previous DMA. Then MSA in current subnet communicates with previous DMA and visit the DMA maintenance table. Based on the DMA maintenance table MSA in current subnet can give a judge of whether switching DMA. MSA can set up a tunnel with DMA or act as a new DMA.

3. Performance analysis

 In MIP-RS approach, when MN moves into a foreign subnet, it joins into a multicast group through the multicast router in the foreign subnet, and then receive multicast packets from the subnet directly. So the multicast path is optimal in MIP-RS approach. In MIP-BT approach, there is a bi-direction tunnel between MN and HA, and MN receive multicast packets through tunnel from HA. However the tunnel may be long. In our approach, DMA transmits multicast packets to MSA of MN by tunnel. Then the MSA transmits the packets to MN. The path between DMA and multicast source is optima and the distance between DMA and MN is limit. Thus it can lessen the problem of triangle route.

For indicating the advantage of the proposed approach in this paper, we compare it with MIP-BT and MIP-RS through the simulation in NS. The main purpose of the simulation is to compare the the length of path of three approaches (hops) and the number of multicast tree restructuring.

First we use nem (network manipulator) to generate six different netowrk topology. There are 400 nodes (router) in each network. Eveary node denote a LAN including a MSA. The detail of the netowrk can be seen in table 1.

Table 1 Parameter in the network topology (1: topology id 2: value 3;parameter)

 

Second we describe how to realize the simulation as follows. We use PIM-SM protocol to set up a multicast tree and select a router as RP in network. Then we select two nodes randomly from all route nodes as HA and initial MSA of MN. The initial MSA is the initial DMA as well. We maintain a buffer table for initial DMA. In the next step, we get all the neighbor nodes around DMA and select one randomly as next MSA of MN. By repeating above process, all passed MSA can be recorded into DMA maintenance table. When the table is filled full, DMA will be changed and a new DMA maintenance table will be built. MIP-RS and MIP-BT approaches are completed as well. At last we record the hops from multicast source to MN and the number of passed subnet before selecting a new DMA.

Third we give the following results and analysis: In each network topology, we select the router node (5) as the RP. The handover times of MN among the subnets is 40 and the size of DMA maintenance table is 5. We complete six simulation through above method and take one simulation as the example. The log information is shown in figure 5.

Figure 5  Log information of one simulation

In figure 5, the node (354) and node (214) is selected as the HA and the initial MSA respectively. In this simulation, the node (0) generate CBR traffic and the total simulation time is 70 seconds. The scenery at the time point t=55s is shown in the figure 6. Where MN has moved to the subnet represented by the node (365) and the RP and BSR are marked by purple. The multicast source and HA are marked by green and gold respectively. The link that MN moved through is red. The multicast traffic and each MSA are all blue. The DMA is marked by yellow.

 

Figure 6  Simulation scenery at t=55s

The length of path (number of hops) of three approaches are shown in table 2 and figure 7. Table 2 denotes the half of total 40 handovers. Where the bold ID denotes the node being DMA.

Table 2 Length of multicast path

 

 

 

 

 

 

 

 

Figure 7 Length of multicast path

 

We complete six similar simulation and compute the average length of multicast path in each simulation. The average length of three approaches i.e. MIT-BT, MIT-RS is described in figure 8.

 

Figure8  Average length of multicast path

 

Figure 9 Ration of path length

 

We compute the ratio of path length in BT (HS) approach to it in RS approach and the ratio of length in DMA approach to it in RS . The ratio is shown in figure 9.

From figure 7, figure 8 and figure 9, we can see the path length in RS approach is shortest. So it is optimal. The path length in BT approach is long. The path length in DMA approach is near it in RS approach and has the sub-optimal path. These results are consistent with the theory analysis.

 Because MN connects the Mbone through MSA directly, RS approach has the shortest path. In this approach, multicast packets can be forwarded through the optimized path, and also avoid the triangle route problem. In BT approach, it cannot conquer the problem of triangle route. As is shown in above figures, there is a long tunnel between the MA and HA. So the path in BT is the longest. In DMA approach, when MSA act as a DMA, MN receives the multicast packets through a optimal path. When MSA communicate with DMA through a tunnel, it will has more hops than RS. But we can control the length of tunnel in a reasonable range and avoid too long tunnel. So the path is near it in RS and has the sub-optimal path.

 At last we will analyze the switch of DMA. The results are shown in figure 10 and figure 11. Figure 10 denotes the selection of DMA and DMA maintenance table. During the 40 times of handover of MN, there is six different DMA. The Y-axis in figure 11 denotes the handover number of MN among subnets and X-axis denotes the new switch of DMA. From the figure 10 we can see the size of DMA maintenance table is 5. Figure 11 denotes the subnet number passed through by MN. When MN moves through more than 5 subnet, a switch of DMA i.e. a multicast tree restructuring will occur. Otherwise in RS approach, there will occur a multicast tree restructuring to each handover of MN. Thus DMA can reduce the restructuring number of multicast tree effectively.

Figure 10 Selection of DMA and DMA maintenance table

 

Figure 11 Handover number of MN among subnets before occurring a new switch of DMA

4. Conclusion

   Supporting multicast service in mobile IPv6 entironment is an important issue in NGI. However the present approach such as MIP-BT and MIP-RS are all have some disadvantage of too long path or frequent multicast tree restructuring. In this paper an approach supporting multicast for mobile node is proposed. The approach presented is based on IETF Bi-directional tunnel and remote subscription. It also integrates the merits of movement based and distance based location management methods. The purpose is to optimize mobile nodes multicast path, and meanwhile reduce the latency and the impact on multicast trees which result from the movement and handoff of mobile nodes. In our approach, between MN and DMA, there is a MSA. MSA communicate with DMA through tunnel and attend multicast mobility management. So the load of MN is low. In addition, the selection of DMA is dynamic and different MN in the same subnet may select different DMA. So the load of one DMA can be lightened. The analysis and simulation denote our approach is novel and efficient. In future we will set up the real testbed to run our approach.

 

References

[1] D. Waitzman, C. Partridge and S. Deering. Distance vector multicast routing protocol. RFC 1075 Novemver 1988

[2] J.Moy. Multicast routing extensions for OSPF. Conmmunications of the ACM 1994;37(8):61-66

[3] D. Fenner, D.Meyer. Multicast Source Discovery Protocol (MSDP). RFC 3618 October 2003

[4] D. Thaler. Border Gateway Multicast Protocol (BGMP). RFC 3913 September 2004

[5] A.Adams, J.Nicholas and W. Siadak. Protocol Independent Multicast-Dense Mode (PIM-DM). RFC 3973 January 2005

[6] S.Bhattacharyya. An Overview of Source Specific Multicast(SSM). RFC 3569 July 2003

[7] Carey L.Williamson, Tim G. Harrison, Wayne L. Mackrell and Richard B. Bunt. Performance evaluation of the MoM mobile multicast protocol. ACM Mobile Network and Applications 1998;3(2):189-201

[8] Suh Y,Shin H, kwon D. An efficient multicast routing protocol in wireless mobile networks. ACM Wireless Networks 2001;7(5):443-453

[9] Y.WANG and W.D CHEN. Supporting IP Multicast for Mobile Hosts. Mobile Network and Applications 2001; 6(1):57-66

[10] Johnson, D., Perkins C., and Arkko, J. Mobility Support in IPv6. RFC 3775 June 2004.

[11] Deering, S., Fenner, W. and B. Haberman. Multicast Listener Discovery (MLD) for IPv6. RFC 2710 October 1999

[12] Zhang, J. Y. Location Management in Cellular Networks. ACM Handbook of wireless networks and mobile computing 2002:27-49

[13] Bar-Noy, A. Kessler, I. and Sidi, M. Mobile Users: To update or not to Update?. Wireless Networks Journal 1995;1(2):175-86.

 

AUTHORS’BIOGRAPHIES

 

Hong-Ke Zhang got his MS degree and PhD degree, majoring in Communication and Information Engineering from University of Electronic Science and Technology of China, Chengdu, People’s Republic of China. Now he is a professor in College of Electronics and Information Engineering, Beijing Jiaotong University, Beijing, People’s Republic of China. His current research interests includes computer communication and router design.

 

 

 

 

 

 

Bing-Yi Zhang is a post-doctor in College of Electronics and Information Engineering, Beijing Jiaotong University, Beijing, People’s Republic of China. He got his PhD degree, majoring in Computer Applications, from the Department of Computer Science, Nanjing University of Science and Technology, People’s Republic

of China. He got his MS degree majoring in Engineering from Henan University of Science and Technology, Luoyang and a BE degree majoring in Engineering in Jiangsu University, Zhenjiang, People’s Republic of China. His current research interests includes network QoS, traffic engineering, mobile wireless IP and signal

analysis.

 

Bo Shen got his PhD degree, majoring in Communication and Information Engineering from College of Electronics and Information Engineering, Beijing Jiaotong University, Beijing, People’s Republic of China. His current research interests includes mobile wireless IP.

 

This work was supported by the National Natural Science Foundation of China under Grant No. 60473001, 60402035