1 / 31

State Analysis and Aggregation for Multicast-based Micro Mobility

State Analysis and Aggregation for Multicast-based Micro Mobility. Ahmed Helmy Electrical Engineering Department University of Southern California helmy@usc.edu http://ceng.usc.edu/~helmy. Outline. Motivation M ulticast-based M obility (M&M) Intra-domain M&M for micro-mobility

darva
Download Presentation

State Analysis and Aggregation for Multicast-based Micro Mobility

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. State Analysis and Aggregation for Multicast-based Micro Mobility Ahmed Helmy Electrical Engineering Department University of Southern California helmy@usc.edu http://ceng.usc.edu/~helmy Ahmed Helmy, USC

  2. Outline • Motivation • Multicast-based Mobility (M&M) • Intra-domain M&M for micro-mobility • Scalability Issues and State Aggregation • Approaches to State Aggregation • prefix vs. bit-wise • perfect vs. leaky • Performance Analysis • Conclusions Ahmed Helmy, USC

  3. Mobile IP - Triangle Routing C Mobile Node (MN) Correspondent Node (CN) B A Home Agent (HA) Ahmed Helmy, USC

  4. Multicast-based Mobility (M&M): Architectural Concept (a) All locations visited by the mobile are part of the distribution tree (at some point) (b) When a mobile moves, only the new location becomes part of the tree - When the mobile moves to a new location, as in (c) and (d) the distribution tree changes to deliver packets to the new location. [A. Helmy, “A Multicast-based Protocol for IP Mobility Support”, ACM NGC ‘00] Ahmed Helmy, USC

  5. Join/Prune dynamics to modify distribution CN CN: Correspondent node (sender) Wireless link Mobile Node Ahmed Helmy, USC

  6. 35000 A+B, Random 30000 A+B, Neighbor 25000 A+B, Cluster 20000 Number of links Average 15000 10000 5000 0 AS r 50 r 100 r 150 r 200 r 250 ARPA ts 50 ts 100 ts 150 ts 200 ts 250 ts 300 ti 1000 ti 5000 ts 1000 Mbone_1 Mbone_2 ts 1008_1 ts 1008_2 ts 1008_3 Topology 16000 C, Random 14000 C, Neighbor 12000 C, Cluster 10000 Number of links Average 8000 6000 4000 2000 0 AS r 50 r 100 r 150 r 200 r 250 ARPA ts 50 ts 100 ts 150 ts 200 ts 250 ts 300 ti 1000 ti 5000 ts 1000 Mbone_1 Mbone_2 ts 1008_1 ts 1008_2 ts 1008_3 Topology Overall Network Overhead Total links traversed. (A + B) /  C = 1.8 Ahmed Helmy, USC

  7. Mean 5 90th percentile Mean 4.5 90th percentile 4 3.5 3 2.5 2 1.5 1 Random Transit-stub Tiers Arpa Mbone AS Topologies (b) Neighbor movement 5 Mean 4.5 90th percentile 4 3.5 Ratio 'r' 3 2.5 2 1.5 1 Random Transit-stub Tiers Arpa Mbone AS Topologies (c) Cluster movement End-to-end Delay 5 4.5 4 3.5 Ratio r=(A+B)/C 3 2.5 2 1.5 1 Random Transit-stub Tiers Arpa Mbone AS Topologies (a) movement Random Ratio ‘r = (A+B)/C’. Average ‘r = 2.11’. Ahmed Helmy, USC

  8. Handoff Latency Ratios Average B/L, C/L and P/L ratios Ahmed Helmy, USC

  9. Conclusion • M&M re-uses many existing multicast mechanisms (simple join/prune) • Extensive simulations show that on average • M&M incurs ~1/2 network overhead as MIP • M&M incurs 1/2 end-to-end delay as MIP • M&M incurs less than 1/2 handoff delay as MIP • M&M outperforms MIP, RO, Seamless HO Ahmed Helmy, USC

  10. Problems with Inter-domain M&M • Requires deployment of inter-domain multicast • Needs global multicast address allocation • State overhead of the multicast tree • Need a new, more practical, approach • M&M for intra-domain micro-mobility Ahmed Helmy, USC

  11. Intra-domain M&M for Micro Mobility M&M BR: Border Router AR: Access Router AP: Access Point Ahmed Helmy, USC

  12. Mobility-proxy Based Architecture (1) Mobile contacts access router (AR) (2) AR sends request to mobility proxy (MP) (3.a) MP performs inter-domain mobility handoff (3.b) MP sends reply to AR with the assigned multicast address Event sequence as the mobile node moves into a domain Ahmed Helmy, USC

  13. Mobility Proxy Mechanisms • MP is dynamically elected and updated (similar to the PIM-SM RP bootstrap problem) • MP keeps mapping for each visiting MN • Another approach is to use algorithmic mapping [on-going work] Ahmed Helmy, USC

  14. Micro Mobility Performance Evaluation and Comparison Topologies: Average # added links: - 2.48; Random Mov - 1.28; Nbr Mov - 1.91; Cluster Mov - 1.89; Overall Av. L for various topologies and movements Ahmed Helmy, USC

  15. M&M vs. Seamless Handoff Previous location, or Seamless handoff (SH) SH/L for various topologies and movements Average SH/L ratio (all topos): - 1.47; Random Mov - 0.84; Nbr Mov - 1.38; Cluster Mov - 1.23; Overall Av. Average SH/L ratio (w/o rand topos): - 1.77; Random Mov - 1.01; Nbr Mov - 1.62; Cluster Mov - 1.47; Overall Av. Ahmed Helmy, USC

  16. M&M vs. Hierarchical MIP Hierarchical MIP of Foreign Agents (FA) FA/L for various topologies and movements Average FA/L ratio (all topos): - 1.51; Random Mov - 3.15; Nbr Mov - 2.06; Cluster Mov - 2.24; Overall Av. Average SH/L ratio (w/o rand topos): - 1.82; Random Mov - 4.61; Nbr Mov - 2.78; Cluster Mov - 3.07; Overall Av. Ahmed Helmy, USC

  17. Comparison Summary • 1080 Simulations (10 per mov/topo/protocol) • In more than 94% of the scenarios M&M outperformed hierarchical and seamless handoff approaches w/o r: without random topologies Ahmed Helmy, USC

  18. Scalability Issues • Scalability of multicast state is still an issue • Unlike unicast, multicast is location independent. • Multicast addresses are not readily aggregatable. Aggregation may not be as intuitive as in unicast • Need a deeper look into multicast aggregation in our architecture Ahmed Helmy, USC

  19. Aggregation Techniques • Prefix Aggregation: • 128.125.50.2 and 128.125.50.3 can be aggregated as one entry as 128.125.50.2/31, where 31 is the mask length • Bit-wise Aggregation: • 128.125.0.2 and 128.125.1.2 may be aggregated as 128.12.0.2\9, where 9 is the position of the aggregated bit. Ahmed Helmy, USC

  20. Aggregation Techs. (contd) • Intuitively bit-wise aggregation gives more chances to aggregate • Deeper look: • sequence of {0,4,1,2,3} leads to 3 states with bit-wise, whereas with Prefix it leads to 2 states • Leaky vs Perfect aggregation • mcast state {S,G,iif, oiflist} or sparse mode {*,G, RP-iff, oiflist} • leaky does not compare the oiflist Ahmed Helmy, USC

  21. Multicast State Aggregation for M&M • Prefix vs. bit-wise Aggregation ratio for in-sequence numbers. Identical gain for bit-wise and prefix aggregation. Ahmed Helmy, USC

  22. Prefix Prefix vs. Bit-wise Aggregation 100 10 Aggregation Ratio Bitwise 1 0 100 200 300 400 500 600 700 800 900 Number of MNs Aggregation ratio for random numbers. Bit-wise aggregation outperforms prefix aggregation up to 80% of the number population. Ahmed Helmy, USC

  23. Multicast State Analysis • Simulations to understand the distribution of state in the nodes and be in a better position to choose the appropriate aggregation using 2 sets of scenarios: • (1) Across space/topology: snapshot of 250k MNs randomly distributed over the topology • (2) Across time: 1000MNs moving 40k moves randomly Ahmed Helmy, USC

  24. BR State Distribution Across Topology: Number of states indexed by the node ID after 250k MNs Ahmed Helmy, USC

  25. Simulated 12 topologies: random, transit-stub, and real networks Obtained consistent results and trends in all simulations Ahmed Helmy, USC

  26. Observations on state distribution across topology • Very clear uneven skewed distribution • Av. state in routers ~ 10k • 80% of nodes had < 10k states • ~ 60% of nodes have around 2.5k states (1% of the total number of MNs). • Heavy concentration in a small number of nodes Ahmed Helmy, USC

  27. 1000 100 State 10 Time 1 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 Node ID State distribution without aggregation • 17-20% of nodes hold more than the average number of states • 40-60% hold less than 1% of the total number of MNs • 66-71% hold less than 2% • That is, we observed a very high concentration of states in only a small fraction of the nodes. State distribution with lossy aggregation Ahmed Helmy, USC

  28. The average aggregation ratio (AR) for the highest 20% of nodes in terms of state was 10.07 (i.e, 90% reduction) • AR of 2 (50% reduction) for average number of states Number of states: Overall average and 90th percentile (w/o agg: without aggregation, w/ agg: with aggregation) • How does aggregation change with # BRs and network routers Ahmed Helmy, USC

  29. Perfect Bit-wise Aggregation BRs Aggregation ratio for perfect aggregation with various topologies and multiple BRs. Ahmed Helmy, USC

  30. Lossy Bit-wise Aggregation BRs Aggregation ratio for lossy aggregation with various topologies and multiple BRs Ahmed Helmy, USC

  31. Conclusions • Aggregation increases with • decrease in number of BRs • increase in number of MNs • decrease in number of network routers • We get better aggregation ratios with concentration of the multicast state • The more concentration, the worse the problem, but the more effective the aggregation • Bit-wise aggregation can reduce state by 90% in nodes with the highest 20% states Ahmed Helmy, USC

More Related