1. Fault-Tolerant Design for Mobile IPv6 Networks Jenn-Wei Lin and Ming-Feng Yang Graduate Institute of Applied Science and Engineering Fu Jen Catholic University

2. 2 2 Outline Introduction The Proposed Approach Simulation Conclusion References

3. 3 3 Introduction Mobile IPv6 Operation AR HA CN Regular IPv6 Router Regular IPv6 Router Home Network Foreign Network Internet Packets MN HA: Home Agent MN: Mobile Node CN: Correspondent Node AR: Access Router

4. 4 4 Introduction Mobile IPv6 Operation AR HA CN Home Network Foreign Network Internet 1. Binding update 2. Binding acknowledgement HA: Home Agent MN: Mobile Node CN: Correspondent Node AR: Access Router Regular IPv6 Router Regular IPv6 Router Handoff MN

5. 5 5 Introduction Mobile IPv6 Operation AR HA CN MN Home Network Foreign Network Internet Packets Tunnel packets Triangle routing problem Handoff HA: Home Agent MN: Mobile Node CN: Correspondent Node AR: Access Router Regular IPv6 Router Regular IPv6 Router

6. 6 6 Introduction Mobile IPv6 Operation AR HA CN MN Home Network Internet 1. Binding update 3. Binding acknowledgement AR Foreign Network Handoff HA: Home Agent MN: Mobile Node CN: Correspondent Node AR: Access Router Regular IPv6 Router Regular IPv6 Router 2. Cache the current location of the MN Route optimization 4. Packets

7. 7 7 Introduction Motivation  If a failure occurs in an HA Its managing MNs (the failure-effected MNs) cannot perform binding updates. If an CN sends a packet to the failure-effected MN but it does not cache the binding information of this MN, the packet will be lost.

8. 8 8 Introduction Goal  Propose an efficient approach to providing fault-tolerant capability for Mobile IPv6 network.  Can handle the lost packets at the faulty HA. Not dependent on the TCP layer to avoid incurring long recovery latency.

9. 9 9 Introduction Network Model A wide-area mobile network → A number of network domains HA: Provide the mobility support for its managing MNs AR: Assist the MNs to forward packets within a network domain. IR: Transmit packets from a network domain to another.

10. 10 Introduction Failure Detection  Detected by MN MN performs a binding update to its serving HA, but it does not receive a binding acknowledgement for a certain period of time.  Detected by CN If a packet to an MN is required to go through the faulty HA, the sending CN will receive an ICMP destination unreachable message.

11. 11 Outline Introduction The Proposed Approach Simulation Conclusion References

12. 12 The Proposed Approach Basic Idea  Initially, each MN gets or updates its preferable external HAs from its default HA.  When an MN detects a failure in its default HA, assign a new serving HA and retrieve lost packets.

13. 13 The Proposed Approach Preferable External HAs

14. 14 The Proposed Approach Preferable External HAs (Cont.)  Initially, each HA stores the identities of other HAs based on the order of their location distances with the HA. The network model of this paper is under a wide-area mobile network, not mobile Internet. The number of HAs, the distance relationship between an HA and others, and the addresses of all the HAs in the system can be known whiling establishing the Mobile IPv6 network.

15. 15 The Proposed Approach Preferable External HAs (Cont.)  The preferable external HAs are the HAs located in the network domains having different distance with the located network domain of the MN Each MN can get its preferable external HAs from its default HA via the binding update message. The new located network domain of the MN can be inferred from the MN’s care-of-address.

16. 16 The Proposed Approach Preferable External HAs (Cont.) Handoff

17. 17 The Proposed Approach Assign a New Serving HA

18. 18 The Proposed Approach Assign a New Serving HA (Cont.) When an MN detects a failure in its default HA Select first (next) preferable external HA as new serving HA Generate a unique external home address Perform a binding update to the first (next) preferable external HA Success Perform a binding update with each communicating CN Yes No In the faulty or overloading status Select the HA that is close to the current location of the MN ( external home address, care-of-address) Route optimization

19. 19 The Proposed Approach Assign a New Serving HA (Cont.) The suffix of the external home address ← The MN’s MAC address MN has a MAC address The prefix of the external home address ← The prefix of the selected external HA The suffix of the external home address ← Use the MD5 hash function to generate Combine the prefix and the suffix of the external home address Combine the prefix and the suffix of the external home address unique Yes No Yes No Generate a unique external home address

20. 20 The Proposed Approach Retrieve Lost Packets

21. 21 The Proposed Approach Retrieve Lost Packets (Cont.)  When an CN would like to send a packet to the failure- effected MN but it has not the MN’s binding information in its binding cache. The packet will be forwarded through the faulty HA, and then lost.

22. 22 The Proposed Approach Retrieve Lost Packets (Cont.)  Before a packet is intercepted by a HA, the packet is first received by a corresponding IR. Therefore, CN will receive an ICMP destination unreachable message from IR. Ask the IR to trace its sending ICMP destination unreachable messages to form an undeliverable packet list. Ask the CN to store the undeliverable packets in its undeliverable packet buffer. 1. Packet 1 3. ICMP destination unreachable

23. 23 The Proposed Approach Retrieve Lost Packets (Cont.)  The recovery process of the undeliverable packets 1.a. Binding update 1.b. Binding update 3. Binding error and Undeliverable packet recovery 2.a. Binding acknowledgement 2.b. Finding the undeliverable packet destined to the MN 4. Solicited binding update 5. Updating the binding information of the MN and finding the undeliverable packets destined to the MN 6. Re-sending the previous undeliverable packet destined to the MN

24. 24 Outline Introduction The Proposed Approach Simulation Conclusion References

25. 25 Simulation Overhead of Tolerating Faulty HA  The information about preferable external HAs is attached on the binding acknowledgement message.  The overhead of the extended binding acknowledgement message : The number of attached external HAs is based on the required fault-tolerant capability of the proposed approach. Allowing 9 HAs to simultaneously fail in the system.

26. 26 Simulation Overhead of Tolerating Faulty HA (Cont.)  Each failure-effected MN can select its preferable external HA as the new serving HA after its default HA fails.  The binding update latency for failure-effected MNs : The preferable external HA is close to the current location of the failure-effected MN. The failure-effected MNs can simultaneously perform binding updates to different new serving HAs.

27. 27 Simulation Overhead of Lost Packet Recovery  The undeliverable packets will be retrieved when a failure-effected MN performs a binding update. The space overheads are mainly determined by the binding update interval. The space overhead introduced by the lost packet recovery : The undeliverable packet listThe undeliverable packet buffer

28. 28 Outline Introduction The Proposed Approach Simulation Conclusion References

29. 29 Conclusion An Efficient Approach to Tolerating the HA Failure in a Mobile IPv6 Network System.  Consider how to recover from the lost packets at the faulty HA.  The simulation results show that the incurred overhead is very small. If The probability that many HAs simultaneously fail is small The binding update interval is not too long,

30. 30 References

31. 31 References

32. 32 Thank you for your listening.