1. 1 A Dynamical Redirection Approach to Enhancing Mobile IP with Fault Tolerance in Cellular Systems Jenn-Wei Lin, Jichiang Tsai, and Chin-Yu Huang IEEE Global Telecommunication Conference 2002 Taipei, Taiwan, R.O.C., Nov. 17-21, 2002 Department of Computer Science & Information Engineering, Fun Jen Catholic University, Taipei, Taiwan jwlin@csie.fju.edu.tw

2. 2 Outline Introduction Introduction Background Proposed Approach Evaluation Conclusions

3. 3 Introduction n Mobility IP in Cellular Systems –Ongoing data sessions without disruption due to mobility –IETF RFC 2002 Two kinds of mobility agents –Foreign agent (FA) –Home agent (HA) –Failure Occurrence Interrupting the data executable capability of mobile users

4. 4 Introduction n Mobile Packet Data Flow –Data request Radio Access Network Mobile Packet Backbone Foreign Agent Home Agent Internet Application Server Mobile Node

5. 5 Introduction n Mobile Packet Data Flow –Data response Home Agent Internet Application Server Mobile Packet Backbone Foreign Agent Radio Access Network Mobile Node

6. 6 Introduction n Failure Occurrence –Failures in FAs Data requests unable to be delivered Home Agent Foreign Agent Internet Application Server Failure Radio Access Network Mobile Node Mobile Packet Backbone

7. 7 Introduction n Failure Occurrence –Failures in HAs Data response unable to be sent back Home Agent Foreign Agent Internet Application Server Failure Radio Access Network Mobile Node Mobile Packet Backbone

8. 8 Introduction n Goal –Not terminating the data services of mobile users when failures occur in mobility agents Proposing a reliable Mobile IP protocol in cellular systems –Tolerating multiple failures of mobility agents –Not needing the hardware support

9. 9 Outline Introduction Background Background Proposed Approach Evaluation Conclusions

10. 10 Background n Wireless Network Model

11. 11 Background n Previous Approaches –R. Ghosh and G. Varghese [1998] –J. H Ahn and C. S. Hwang [2001] –Features Mobility agent –Hardware replication Mobility information –Potential long registration delay –Stable storage Fault-tolerant range –Within a network segment

12. 12 Outline Introduction Background Proposed Approach Proposed Approach Evaluation Conclusions

13. 13 Proposed Approach n Basic Idea –Workload redirection Network-initiated handoff –Redirecting the workload of the faulty FA to other failure-free FAs Tunneling –Redirecting the workloads of the faulty HA to other failure-free HAs

14. 14 Proposed Approach n Network-Initiated Handoff –Modifying the FA selection algorithm Relationship between RANs and FAs before a FA failure

15. 15 Proposed Approach n Network-Initiated Handoff –Resetting the FA selection algorithm (Cont.) Relationship between RANs and FAs after a FA failure Virtually moving the locations of MNs under the coverage area of the faulty FA

16. 16 Proposed Approach n Tunneling –Performing the tunneling on the neighbor routers

17. 17 Proposed Approach n Tunneling –Intercepting the response packets by failure- free HAs

18. 18 Proposed Approach n Reconstructing mobility information –Sending a mobility-reconstruction message to each FA –Filtering the visitor list to find the MNs originally managed by the faulty HA –Re-organizing each selected visitor entry as the form of a mobility binding entry

19. 19 Outline Introduction Background Proposed Approach Evaluation Evaluation Conclusions

20. 20 Evaluation n Performance Degradation on a Failure-Free Mobility Agent –Probability of blocking packet data in a failure-free mobility agent –Packet data to a mobility agent Poisson distribution –Processing time of packet data in a mobility agent Arbitrary distribution –M/G/c/c queuing model

21. 21 Evaluation n Blocking Probability of a Failure-Free Mobility Agent –Erlang’s loss formula given from the M/G/c/c queuing model

22. 22 Evaluation n Traffic Parameters ParameterMeaningValue F Number of faulty mobility agents1 c Number of resource units in a mobility agent 50 (100) λ Arrival rate of data to a mobility agent 10, 25, 50, 100 μ Service rate of data in a mobility agent 1 wkwk Ratio for redirecting the workload of the faulty mobility agent to the failure-free mobility agent k 0.1 to 1.0

23. 23 Evaluation n Blocking Probability of a failure-free FA

24. 24 Evaluation n Blocking Probability of a failure-free HA

25. 25 Evaluation n Blocking Probability –When is not too large (e.g. 10 in FA and 10, 25 in HA), the blocking probability nearly approaches 0 regardless of the variance of w k. –When is very large (e.g. 100), the blocking probability may be not large for the smaller w k. –Four used traffic intensities are larger than the general traffic intensity in a commercial wireless system

26. 26 Outline Introduction Background Proposed Approach Evaluation Conclusions Conclusions

27. 27 Conclusions n A New Approach to Tolerating Multiple Failures of Mobility Agents –Not incurring failure-free overhead –Not requiring hardware support –Dynamically generating the backups of faulty mobility agents n Overhead –Performance degradation on a failure-free mobility agent –M/G/c/c queuing model (Erlang’s loss formula)

28. 28 Conclusions n Comparisons Comparing Metrics Proposed Approach in Ghosh and Varghese [1998] Approach in Ahn and Hwang [2001] Hardware support NoYes Range Whole system network In a network segment Failure-free overhead No Potential long registration Message logging and checkpointing Failure overhead Network- initiated Handoff Tunneling Searching ARP execution Restoration

29. 29 Thank You for Your Attention Jenn-Wei Lin jwlin@csie.fju.edu.tw