An Active Network Based Hierarchical Mobile Internet Protocol Version 6 Framework

Transcription

1 An Active Network Based Hierarchical Mobile Internet Protocol Version 6 Framework Zutao Zhu Zhenjun Li YunYong Duan Department of Business Support Department of Computer Science Department of Business Support Guangdong Telecom Shenzhen Branch Sun Yat-sen (Zhongshan) University Guangdong Telecom Shenzhen Branch Shenzhen, CHINA Guangzhou, CHINA Shenzhen, CHINA zzt@sztelecom.com jordanli_flying@hotmail.com yyduan@sztelecom.com Abstract The Mobile Internet Protocol version 6 (MIPv6), based on IPv6, will be widely used in the next generation wireless network. The latency of the handover is one of the most important parameters in evaluating MIPv6. At present, the popular wireless handover technologies include the Fast Mobile Internet Protocol version 6 (FMIPv6) and the Hierarchical Mobile IPv6 (HMIPv6) invented by IETF and the Active Network invented by DARPA. In this paper, we propose a new strategy of the wireless handover in MIPv6, which is based on the HMIPv6 and the Active Network. The strategy will optimize the handover process during MN s roaming. In the Network Simulator (NS-2), we set up a simulation scenario. We compare the handover performance of the network architecture using our strategy with that of HMIPv6 and prove that our strategy has better performance. Keywords-HMIP; active network; QoS; MIPv6 I. INTRODUCTION With the development of the Mobile Internet Protocol version 6 drafts, the demands of QoS in the Mobile IPv6 network environment are increasing. Nowadays, researches are focused on [6]: (1) QoS support architecture in the Mobile IPv6 network environment. The architecture addresses how to deploy each network node and each network component node and how to communicate with each other. (2) Authentic Authority Accounting (AAA) over Mobile IPv6. It covers the authentication of Mobile Nodes (MN), access control, which includes part of QoS control, and accounting policy. It is widely believed that QoS of MIPv6 can be improved through modifying corresponding signaling. (3) Handover of Mobile IPv6 [2]. This is the hotspot of mobile QoS research and the key for improving QoS. In the latest research of IETF Mobile IP Group, there are two kinds of mobility: Macro-mobility and Micro-mobility. The latter happens frequently and the moving region is small. IETF has some micro-mobility architecture drafts. Two of the most famous are HMIPv6 [1] and FMIPv6. In HMIPv6, a new node called Mobility Anchor Point (MAP) is introduced. A MAP is essentially a local Home Agent (HA). The MAP limits the amount of Mobile IPv6 signaling outside the local domain. Active networks, proposed by DARPA in 1994, allow individual user, or groups of users, to inject customized programs into the nodes of the network. "Active" architectures enable a massive increase in the complexity and customization of the computation that is performed within the network, e.g., that is interposed between the communicating end points. Active networks are a novel approach to network architecture in which the switches of the network perform customized computations on the messages flowing through them. In an active network, the routers or switches of the network perform customized computations on the messages flowing through them [5]. II. ACTIVE NETWORK BASED HMIPV6 FRAMEWORK A. Overview We propose a HMIPv6 framework which is based on active network and HMIPv6. The framework obeys the basic QoS design principle, separate principle. The mobility management solution is using optimized HMIPv6. We add a new component, the active network, into the framework. With the help of the active network, the QoS of mobile node can be significantly

2 3Com IBM d i g i t a l 3Com imac improved. The network architecture shown in Figure 1 illustrates an example of the use of the active router in a visited network. B. Basic Components The mobility management of our framework is based on the latest HMIPv6 draft of IETF of December 9 th, As we know that HMIPv6 has the shortcoming of sacrificing routing optimization, we modify HMIPv6 so that it can optimize routing. Figure 1. An HMIPv6 Framework Using the Active Network According to the HMIPv6 model, the components such as Mobile Node (MN), MAP, Access Router (Active Router), Core Router (CR), QoS Proxy, AAA, Home Agent (HA), Correspondent Node (CN) and IPv6 network are necessary. By using the active network, the framework can optimize the handover of MN. In order to use a active network, active routers are needed. Besides the function of the traditional router, the active router strengthens the handle of active packets. An active router is a programmable device and active programs are stored in the active router. Active routers maintain the information of MN, such as the address of HA, the Regional Care-of Address (RCoA), the on-link Care-of Address (LCoA) and the address of CN. Active routers track and handle information that comes from a MN and a CN so that it can get the updated information. If there is no updated information in a life cycle, the active router will delete the related information. C. Deployment and Interaction Our framework involving the active network and HMIPv6 is shown in Figure 1. In the framework, there are at least one MAP, one AAA, one QoS Proxy, one CR, one HA, one CN, one MN and several active routers. The most significant improvement of our framework is on the mobility performance. We concentrate on the optimization of handover involving the active network. When a MN moves to a new Access Router (nar), with fast handover being used, the old Access Router (oar) broadcasts the MN s network interface information to the router. Subsequently, the active program actively generates the CoA of the MN, in our framework, that is, LCoA and monitors network neighbor requests and neighbor broadcast messages and detects whether there is a duplicate address. If the selected address has not been occupied, the active router sends an option message containing the flag which indicates that the Duplicate Address Detection (DAD) has been performed. When receiving the option message, the MN can use the address immediately and doesn t need to detect address confliction. In this manner, we optimize the fast handover processing. When a MN sends Binding Updates (BU) message to HA and CN, the active router catches the message, recording the latest information of the MN and tracking the mobility of MN, and sends Binding Success message to the MN at the same time. It can reduce the Round Trip Time (RTT). Then, the active network forwards the BU message to the HA and the CN respectively. It is obvious that the handover latency depends on not the RTT between the new position of the MN and the HA but on the RTT between the MN and the nearest active router loading the active program. The active router is responsible to forward packets resulting from the handover of the MN. The active router forwards packets destined to the old Care-of Address (ocoa) of the MN to the MN s new Care-of Address (ncoa). When the MN frequently performs handoffs, the active route saves the packets which are sent to the MN, avoiding the overflow which happens in a legacy router. In the framework, upon arrival in a visited network, receiving active router s advertisement containing information on one or more local MAPs, the mobile node can bind its current location (on-link CoA) with an address on the active router s subnet (RCoA) with the help of the MAP and the active router. A MN sends local Binding Updates to the MAP via the active router. After the active router returns a Binding Acknowledgement indicating a successful registration to the MN, the mobile node starts to use RCoA and LCoA. At the

3 same time, the active router sends binding updates which has QoS option header to the HA and the CN via tunnels. The MAP updates the binding cache entry with MN s RCoA and LCoA and sends update signaling to the Core Router and the active routers within the domain so that the Core Router can forward the packets addressed to the MN directly to the active routers accessed by the MN, instead of forwarding by the MAP. By this means, we overcome the shortcoming of the MAP s overload which is not capable of routing optimization in HMIPv6. In this manner, network data transmission is separated from signaling control and the MAP is concentrated on network management and control. Involving the active router, MN operations have been extended and optimized. If mobile node and correspondent node are in the same domain, according to the HMIPv6 specification, the mobile node needs to register to the MAP and communicate to the correspondent node by tunnel. Compared with direct communication, it leads to more delay and overload. We make an improvement in using the active router. The active router detects whether the correspondent node communicated with the mobile node belongs to the same domain. If the mobile node and the correspondent node belong to the same network, the packets sent to CN are routed directly to the access point of the CN. If not, the packets are forwarded by the Core Router. We will use pseudo code [7] to demonstrate how the main components operate when they communicate in active router based HMIPv6 framework. These main components include the Mobile Node, the MAP and the Core Router. (1) Mobile Node Operations In our framework, the function of the MN has been simplified. Because of the involving of the active router, the MN only needs to obtain its RCoA and LCoA according to the Router Advertisement in the case of handover between different administrative domains. The MN updates its LCoA when handover takes place intra-domain. Algorithm 1. Mobile Node mobility management 1: Loop /*wait for event*/ 2: If a new MAP domain is entered (receiving RA containing MAP option) then 3: form new LCoA and RCoA based on RA 4: send local binding updates to access active router 5: if the active router returns an address confliction message then 6: go to step 3 7: else use new LCoA and RCoA 8: End if 9: End if 10: If access point is switched (receiving RA containing new routing option) then 11: form new LCoA with access active router 12: End if 13: End loop (2) MAP Operations The MAP is in the core position of the HMIPv6 mobility management. It is responsible for receiving and handling packets sent by the Core Router and forwarding them to the mobile node. Binding updates are also tunneled to the HA and the CN by the MAP when a MN handover takes place. In our framework, the forwarding function related to the MN of the MAP is implemented in the active router, so the MAP mainly handles the communication management. The MAP sends Binding Updates to the Core Router when a MN arrives at a visited network. The Core Router then redirects the routing path to the MN. When receiving the Binding Updates of the MN, the MAP not only updates its Binding Cache, but also sends update messages to the active routers in the same domain so that these active routers obtain the MN s new access interface. When the Core Router receives packets destined to unknown destination, it forwards to the MAP by default. Algorithm 2. MAP handling 1: Loop 2: If receive a Binding Updates message then 3: If the RCoA doesn t exist in Cache then 4: record the MN s new RCoA and LCoA and network address of active router 5: For every active router in the same domain 6: send the MN s new RCoA and LCoA and the address of the access router 7: End for 8: send MN s new RCoA and LCoA to CR 9: Else 10: send MN s LCoA to each active router and CR 11: End if

4 12: Else /*receiving packets addressed to local network or no entry in CR*/ 13: If no entry is in the CR then 14: save the data in case of data stack non-overflow 15: Else /*destination address is in the local network*/ 16: forward packets according to the path in Cache 17: End if 18: End loop (3) Active Router Operations As an implementation of the active network in our framework, the active router should detect whether the MN s RCoA and LCoA are conflicting and send Binding Acknowledgement message to the MN after receiving the Binding Update message sent by the MN to the HA and the CN so that the MN can use the new address. The active router is responsible to forward the MN s packets, directed inter-domain or intra-domain. Algorithm 3. Active router handling 1: Loop 2: analysis address occupation via RAs 3: If receive MN s address request message then 4: If the address has been occupied then 5: return address occupied message 6: Else 7: return address available message 8: End if 9: End if 10: If receive BU from MN then 11: return binding success message to the MN 12: send BUs to the HA and the CN 13: send the MN s RCoA and LCoA to the MAP 14: End if 15: If receive routing update message from the MAP (notifying MN s new LCoA) then 16: update route table 17: End if 18: If the MN send packets addressed to in-domain address then 19: directly address packets to the corresponding active router 20: Else 21: forward to Core Router 22: End if 23: End loop III. SIMULATION AND COMPARISON A. Simulation Setup The studied scenario is designed in order to be large enough to provide realistic results but to be small enough to be handled efficiently within NS-2[3]. The chosen scenario, depicted in Figure 1, is composed of the HA and the CN that are connected via the Internet, actually an IPv6 network, to a central router (CR). Four access routers (AR) or active routers (in our architecture), each one representing a different IP subnet, are connected via a MAP to the CR. When the simulation starts, the mobile nodes are uniformly distributed over the coverage area. Figure 2. Topology in simulation The access routers or active routers have been positioned in a way to provide total coverage to an area of approximately square meters considering a transmission range of 600 meters, as shown in Figure 1. The coordinates of ARs are (200,800), (800,800), (200,200), (800,200), respectively. The mobile nodes move randomly within the coverage area following the random waypoint mobility model. The bandwidth of wireless part is 2Mbps. The Internet, connected with the central router, HA and CNs, is modeled as a 5Mbps duplex link with a default Link Delay (ld) of 100ms. In the simulation, the ld of MAP is 5ms and the ld of HA and CN is 10ms respectively. The number of MNs is increases by 5. As a TCP Constant Bit Rate (CBR) sources, which provide constant traffic where no acknowledgments are required, CN communicates with MN in 88kbps. MN receives packets from the shared AR queue and MN also competes with other MNs

5 and with an AR to access the channel, half of the MNs receiving data from the CNs and the other half sending data to the CNs [4]. In this paper, we define the interval from CN s sending out a packet before MN s handover to MN s receiving the packet after handover as the y-axis (delay). MN s roaming takes place 5 seconds later after the simulation starts. We can get a result with a special number of MN after 100 times simulations. Then we calculate the packet loss in this simulation. The results of this simulation are shown in the Figure 2. the latency of MN-CR(AR) are 35ms, 55ms and 90ms, respectively. If ignoring the time-consuming in the active routers, we can make a conclusion that the handover latency in our architecture is 30% optimizing than that in the HMIPv6, especially when the number of MNs is increasing. The number of dropped packets is 25% optimizing too. The conclusion of simulation is that efficiency of handover in our architecture is better than that in HMIPv6. IV. CONCLUSIONS We propose our HMIPv6 framework using the active router and state the operations of each component in detail and describe how to handle the handover of the MN. In the end, through simulation comparison and theoretical analysis, we draw a conclusion that our framework can significantly decrease the latency when handover happens. Figure 3. Simulation Result B. Analysis of Simulation Results In HMIPv6, the latency of MN handover is 165ms, 200ms and 265ms when the number of MNs is 30, 40 and 50, respectively. And the number of loss packets of MN is 1.5, 2.0 and 3.0 when the number of MNs is 30, 40 and 50, respectively. In our architecture, the latency of MN handover is 150ms, 170ms and 205ms when the number of MN is 30, 40 and 50, respectively. And the number of loss packets of MN is the number of MN is 1.2, 1.5 and 2.2 when the number of MNs is 30, 40 and 50, respectively. Since the handovers in our simulation are all In-Domain-roaming, MNs are not necessary to send BU to HA. The latency of handover should ignore the internet delay, 100ms, and the ld of CN to Internet, 10ms and the Internet to CR(5ms). In all, we subtract 115ms from the latency. So we get it as this: in HMIPv6, the latency of MN-CR(AR) is 50ms,85ms and 150ms, respectively while in our architecture, ACKNOWLEGEMENT The authors are very grateful to Prof. Cheng for his careful readings of earlier versions of this paper and insightful discussions. The authors would also like to thank the reviewers for their helpful comments and suggestions. REFERENCES [1] Hesham Soliman, Claude Castelluccia, Karim El-Malki, Ludovic Bellier. Hierarchical Mobile IPv6 mobility management (HMIPv6).IETF Draft.December, [2] Hee Young Jung, Jun Seob Lee, Seok Joo Koh, Hesham Soliman, Karim El-Malki.Fast Handover for Hierarchical MIPv6 (F-HMIPv6).IETF Draft, May,2003. [3] NS-2 resource. [4] M.Torrent-Moreno. A Performance Study of Fast Handovers for Mobile IPv6. IEEE International Conference on Local Computer Networks, 2003 [5] David L. Tennenhouse, et. al., A Survey of Active Network Research. IEEE Communications Magazine, Vol. 35, No.1, pp January 1997 [6] Liao Xiaofei et. al., Mobile IPv6 Handoff Scheme Based On Active Network Techology, [7] Yang Chen. New QoS Architecture in Mobile IPv6 (2003), Master Thesis.Univ. of Sci. & Tech. Of China.