1. Smart Routers for Cross-Layer Integrated Mobility and Service Management in Mobile IPv6 Systems Authors: Ding-Chau Wang. Weiping He. Ing-Ray Chen Presented by: Dong Chen. Zhiqian Chen. Brad Herald

2. Overview 1.Introduction Internet Protocols (IP); IPv4, and IPv6 Mobile IP; MIPv6 and HMIPv6 2.DMAPwSR with Smart Routers; Dynamic Mobility Anchor Points with Smart Routers 3.Performance Model 4.Numerical Results with Simulation Validation 1.Comparison of DMAPwSR and HMIPv6 2.Simulation Validation 5.Conclusion 2

3. 1. IP protocols Internet Protocol version 4 (IPv4) is a connectionless protocol for use in packet-switched networks. It delivers packets often out of order and usually duplicated. IPv4 carries more than 96% of the worldwide Internet traffic. Internet Protocol version 6 (IPv6) is the latest IP version, RFC 2460, December 1998. Improves addressing networks, part of an IPv6 address contains a host configurable address. Improves network-layer security. Data portion of packet can be as large as 2 32 -1 octets. IPv4 addresses are 32 bits long and number about 4.3×10 9 (4.3 billion) IPv6 addresses are 128 bits long and number about 3.4×10 38 (340 undecillion) 3

4. Mobile IP Usage MIPv6 – with advances in wireless IP-based networks and increased numbers of wireless devices, it is speculated MIPv6 will become more prevalent in the next generation of all-IP networks. This will allow users to have continuous service while on the go. HMIPv6 – The hierarchical version of MIPv6 is designed to reduce the network signaling cost for mobility management based on the observation that statistically local mobility accounts for more than 60% of movements made by a MN. 4 MIPv6: Mobile IPv6 HMIPv6: Hierarchical MIPv6 MN: mobile node CoA: care-of-address RCoA: regional CoA CN: corresponding nodes AR: access routers SMR: service to mobility ratio HA: home agent SPN:stochastic Petri net Terminology

5. 2. DMAPwSR With Smart Routers What is DMAPwSR? Basic Idea Rationale 5

6. Mobility and Service Management Scenarios under DMAPwSR 6 He, Weiping, Ing-Ray Chen, and Ding-Chau Wang. "Cross-layer integrated mobility and service management utilizing smart routers in mobile IPv6 systems." Proceedings of the 8th International Conference on Advances in Mobile Computing and Multimedia. ACM, 2010.

7. Issue DMAP size Small area: DMAP will change more often Service delivery cost is low for communication between CN and MN Location management cost is high, more frequent updates for RCoA Large area: DMAP will not change as often Service delivery cost is high for communication between CN and MN Location management cost is low, infrequent updates for RCoA 7

8. 3. Performance Model Use SPN to model a MN’s mobility behavior in HMIPv6 and DMAP to find the optimal DMAP service area size. SPN can model the extensive number of states and a MN’s behavior migrating from one state to another state in response to events occurring in the system. The goal is to find the minimal mobility and service operations cost per time unit. Communication Cost includes Mobility management overhead Update DMAP with CoA changes Informing HA and CNs of RCoA changes Service management overhead to deliver data packets between MN and CNs 8 Data packet rate between the MN and CNs  Mobility rate at which the MN moves across boundaries SMRService to mobility ratio ( /  ) NNumber of server engaged by the MN KNumber of subnets in a DMAP service area  1-hop communication delay per packet in wired networks  Average distance between HA and DMAP  Average distance between CN and DMAP  Cost ratio, wireless versus wired

9. Performance Model based on SPN Transitions Timed - Moves, MN2DMAP, and NewDMAP Immediate – A and B Tokens – represents an event occurrence Places – Moves,Intra,and Xs Arcs 9 Transition Move occurs at rate , mobility rate for MN crossing boundaries. Place Moves holds a token when a subnet crossing event happens. MN2DMAP transition represents the MN registering with DMAP the new CoA received from a new AR. For K - ARs: Intra-domain (K-1) - each token represents a MN moving to each AR within the domain Inter-domain (Kth move) – last MN move in domain triggers transition to NewDMAP Place Xs holds number of subnet crossing events since last DMAP registration.

10. Communication Cost 10

11. Communication Cost 11

12. Communication Cost 12

13. 4. Numerical Results with Simulation Validation Tools SPN Model for Analysis SMPL for Simulation DMAPwSR is movement-based DMAP service area size determined by the # of movements the MN DMAP service area size could be distance-based, and it is the distance between the DMAP and the current subnet Compare simulation result of movement-based versus distance-based to see if it is sensitive to definition of DMAP service area 13

14. Network Coverage Models Compare simulation result of the network coverage model used to see if it is sensitive. 2D Hexagonal-shape network coverage Mesh network coverage Based on real trace data 14

15. 2D Hexagonal-shape Network Coverage Model Moves from current AR to one of the 6 neighbor ARs randomly with equal probability of ⅙ Can be used for both movement/distance based Subnet area is represented by radius r Distance-based DMAP area service will have (2K-1) r K is the DMAP service area size 15

16. Mesh Network Coverage Model moves from current AR to one of the 6 neighbor ARs randomly with equal probability of ¼ 16

17. Trace-Data Based Network Coverage Model two AP are neighbor APs if they are separated in distance in the range of [100,200m]. (AP signal coverage range is about 91.4m) MN randomly choose a neighbor AP when leaves the current AP. 17

18. 4.1 Comparison of DMAPwSR with MIPv6 and HMIPv6 18

19. Comparison of DMAPwSR with MIPv6 and HMIPv6 19 Higher SMR means the MN’s packet arrival rate is much higher than the mobility rate. Comparing cost: MIPv6 and DMAPwSR Low SMR, DMAPwSR dominates over MIPv6. As SMR increases, in the case 64, K opt approaches 1 which DMAPwSR is essentially the same as basic MIPv6. Comparing cost: HMIPv6 and DMAPwSR As SMR increases, HMIPv6 cost decreases compared to DMAPwSR until K opt coincides with K H. After K H HMIPv6 cost increases sharply. DMAPwSR performs better than HMIPv6 with low or high SMR.

20. Comparison of DMAPwSR with MIPv6 and HMIPv6 Low SMR, the cost gain of DMAPwSR over HMIPv6 can go as high as 40%. High SMR, the cost gain is less than 5%. The cost gain is per-MN per time unit. Over time, a 5% cumulative gain over all MNs will still be significant. 20

21. Comparison of DMAPwSR and HMIPv6 (with different  and  values) The cost differences are more pronounced as the distance between the HA (or CN) and the MN increases. 21

22. 4.2 Simulation Validation Batch Means Analysis Technique Simulation period is divided into batch runs Each consists of 2,000 observations Minimum batch size is 10 Target metric: grand mean of overall network cost Additional batches are needed if grand mean is within 95% confidence level and 10% accuracy from the true mean. To achieve confidence level of 95% and accuracy of 5%, normally takes more than 20,000 observations. 22

23. Service Area Size K vs Total Cost This figure shows there exists K that can balance the trade-off between large service delivery cost using large service area versus large location management cost of informing HA and CNs using small service area. 23

24. Cost Difference vs SMR (by simulation versus analytical results) 24

25. Cost Difference vs SMR (by movement-based versus distance-based) 25

26. Cost Difference vs SMR (under various time distributions) Normal, uniform, and exponential distribution the results trend remain the same no matter which time distribution is. 26

27. K vs SMR (under various time distributions) Normal, uniform, and exponential distribution the results trend remain the same no matter which time distribution is. 27

28. Cost Difference vs SMR (under various network coverage models) Hexagonal, mesh, trace-data based remarkably consistent with each other. 28

29. Summarizing Results The result is insensitive: analytical vs. simulation movement vs. distance based various time distributions various network coverage models 29

30. 5. Conclusion DMAPwSR SPN DMAPwSR Model Simulation Validation Future work 30