Adaptive Load Balancing based on IP Fast Reroute to Avoid Congestion Hot-spots

1. Adaptive Load Balancing based on IP Fast Reroute to Avoid Congestion Hot-spots Masaki Hara, Takuya Yoshihiro 2011 IEEE International Conference on Communications (ICC)

2. Outline • Introduction • Load balancing mechanisms using IP fast reroute scheme SBR • Traffic simulation • Concluding remarks

3. Introduction • Introduction • Load balancing mechanisms using IP fast reroute scheme SBR • Traffic simulation • Concluding remarks

4. Introduction (cont.) • Recent growth of the Internet and the rapid increase of user traffic require more bandwidth and communication quality for infrastructure networks. To make the most of existing network resources, several traffic engineering (TE) techniques have been proposed so far. • Traffic engineering • IP based TE • MPLS based TE

5. Introduction (cont.) • Multiprotocol Label Switching (MPLS) • Reduces network complexity • Compatible with existing network technology • Reduce network cost • To ensure QoS and security • Traffic engineering capabilities Expensive

6. [2] B. Forts and M. Thorup, “Internet Traffic Engineering by Optimizing OSPF Weights,” in Proc. INFOCOM2000, pp.519–528, 2000 • First presented the method to optimize link metrics. They formulate the link metric optimization problem by means of linear programming which compute the optimal link metrics to make the most of network resources assuming that traffic demand among every pair of nodes is given. • Their work also presented that IP based TE achieves the similar level of load balance performance compared to MPLS based TE techniques.

7. Introduction (cont.) • In IP based shortest-path scheme, traffic tends to be concentrated on particular links or nodes since the shortest-paths tend to use the common low-cost links.

8. [3] S. Dasgupta, J.C.de Oliveira, and J.P. Vasseur, “A Performance Study of IP and MPLS Traffic Engineering Techniques under Traffic Variations,” IEEE Globecom2007, pp.2757-2762, 2007. • Their simulation demonstrated that under the traffic variation coming from link failure, IP based TE often brings severe congestion hotspots caused of concentration of shortest paths, while MPLS based TE cause much less congestion. • This result implies that there is still room to improve IP based TE techniques.

9. [4] A.K. Mishra and A. Sahoo, “S-OSPF: A Traffic Engineering Solution for OSPF based Best Effort Networks,” In Proc. IEEE Globecom 2007, pp.1845–1849, 2007.[5]M. Anti´c and A. Smiljani´c, “Oblivious Routing Scheme Using Load Balancing Over Shortest Paths,” In Proc. IEEE ICC 2008, pp.5783–5737, 2008. • In practice, the optimization process requires the mechanisms to measure the traffic demand, to share the measured information among all nodes, and to re-compute the new distribution paths. This process would results in losing sensitivity for dynamic transition of traffic.

10. Introduction (cont.)

11. IP Fast Reroute • In this paper, as another approach of load balancing in IP routing schemes, we propose an IP fast reroute based load balancing mechanisms

12. The approach of load balancingusing IP fast reroute has several good characteristics :

13. Introduction (cont.) • This work is the first specific mechanism description and the first investigation on the performance of IP fast reroute based load balancing mechanisms.

14. Load Balancing Mechanisms Using IP Fast Reroute Scheme SBR • Introduction • Load balancing mechanisms using IP fast reroute scheme SBR • Traffic simulation • Concluding remarks

15. Single Backup-table Rerouting • SBR (Single Backup-table Rerouting) • SBR is an IP fast reroute scheme which recovers every single link failure with low overhead, i.e., two 1-bit flags in packet header and one additional routing table. • Two 1-bit flags • b-flag(1-bit) : 0/1 • r-flag(1-bit) • One additional routing table=Tworouting table • primary routing table • backup table (backup/ switch)

16. Single Backup-table Rerouting(cont.)

17. Single Backup-table Rerouting(cont.) packet (b-flag, r-flag) u v w (0, n/a) x y destination d

18. Single Backup-table Rerouting(cont.) packet (b-flag, r-flag) u v w (0, n/a) x y destination d

19. Single Backup-table Rerouting(cont.) packet (b-flag, r-flag) u v w (0, n/a) x y Jam destination d

20. Single Backup-table Rerouting(cont.) packet (b-flag, r-flag) u v w (1, n/a) (0, val.) (0, n/a) u → x → u → v → w → y → d x y Jam d destination

21. Single Backup-table Rerouting(cont.) packet (b-flag, r-flag) u v w (1, n/a) (0, val.) (0, n/a) Jam x y Jam destination d

22. The Proposed Load Balancing Mechanisms

23. Original Router Router Output queue To backup next-hop Incoming packets Router decision using b-flag and r-flag Output queue To primary next-hop

24. Proposed Router q(b) Router Output queue To backup next-hop Incoming packets Router decision using T1 and T2 T1 T2 Output queue To primary next-hop T2 T1 q(p)

25. Original flag packet (b-flag, r-flag) u v w (0, n/a) x y destination d

26. Proposed flag packet (b-flag, r-count) u v w (0, n/a) x y destination d

27. Formal Forwarding Process at Each Router in The Proposed Method

28. Single Backup-table Rerouting(cont.) packet (b-flag, r-count) u v w (0, 2) (0, 1) (1, 1) (1, 0) (0, 0) MAX r-count = 3 Jam x y Jam destination d

29. Traffic Simulation • Introduction • Load balancing mechanisms using IP fast reroute scheme SBR • Traffic simulation • Concluding remarks

30. Simulation Setup • ns-2 simulator • Network topology of Waxman model with 30 and 100 nodes • topology generator BRITE • 10% of the links randomly as “bottleneck” links which bandwidth is 50Mbps • normal link bandwidth is 100 Mbps • 10Mbps CBR flows with 1-kilobyte packets • source and destination nodes are selected randomly • Output queue length is 50 packets. • Threshold T1 is 100% and T2 is 10% • The number of maximum allowed detouring is set to 3. • Simulation is done by generating 10 topologies • for each topology we select different two sets of bottleneck links and traffic patterns • average values to show the results

31. Fig. 3. Improvement of Throughput with Variation of Traffic Load

32. Fig. 4. Ratio of Packets Rescued by Detouring

33. Fig. 5. The Specification of Number of Detour to Reach Destination (Case of 30 Nodes)

34. Fig. 6. The Specification of Number of Detour to Reach Destination (Case of 100 Nodes)

35. Fig. 7. Increase of Packet Delay of Shortest-Path Traffic

36. Fig. 8. Ratio of Throughput of Shortest-Path Traffic

37. Concluding Remarks • Introduction • Load balancing mechanisms using IP fast reroute scheme SBR • Traffic simulation • Concluding remarks

38. Concluding Remarks (cont.) • In this paper we investigated a load balancing technique based on IP fast reroute mechanisms. Our method detours packets only when the packets meet congestion to make the most of vacant resources of networks.

39. Weekly Sentence • Note that since the number of flows transmitted is the same, the amount of shortest-path traffic is the same in both OSPF and the proposed.