1 / 19

A Study of Packet Delivery Performance during Routing Convergence

A Study of Packet Delivery Performance during Routing Convergence. Dan Pei, Lan Wang, Lixia Zhang, UCLA Dan Massey, USC/ISI S. Felix Wu, UC Davis. Packet Delivery during Routing Convergence. D. F. G. A. B. C. E. Failures do occur in the Internet

adie
Download Presentation

A Study of Packet Delivery Performance during Routing Convergence

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. A Study of Packet Delivery Performance during Routing Convergence Dan Pei, Lan Wang, Lixia Zhang, UCLA Dan Massey, USC/ISI S. Felix Wu, UC Davis

  2. Packet Delivery during Routing Convergence D F G A B C E • Failures do occur in the Internet • 20% of intra-ISP links have a MTTF < 1 day [Diot:IMW02] • 40% of Inter-ISP routes have a MTT-Change < 1 day [Labovitz:FTCS-29] • Routing convergence after failure takes time • IS-IS(Intra-ISP protocol): 5+ seconds [Diot:IMW02] • BGP(Inter-ISP protocol): 3+ minutes [Labovitz:Sigcomm00] • Packets can be delivered during convergence

  3. Goal of this paper • How to maximize packet delivery during routing convergence? • Previous work focused on: preventing loops, minimizing convergence time and routing overhead • Topological connectivity’s impact? • Studying: RIP, Distributed Bellman-Ford(DBF), BGP • This problem becomes more important with • Larger Internet topology [Huston01] --> higher freq. of component failures • Richer connectivity[Huston01] --> potentially helps with more alternate paths • Higher bandwidth --> more packets sent during convergence

  4. Outline for the rest of the talk • Introduction of RIP, DBF and BGP • Simulation results and lessons learned • Conclusion

  5. Protocols Examined (I):RIP and DBF Distributed Bellman-Ford(DBF) • Keep distance info from all neighbors D D:3 D:2 D:1 Both RIP and DBF: F • 30sec refreshing interval • Damping timer to space out two triggered updates: 1~5 seconds • Poison reverse: B sends infinity distance to A D:2 A B C E D: infinity D:3 RIP • Exchange distance info. • B’s route to D: Nexthop=A, Dist=4 • B’s route to D: Nexthop=A, dist=4 Alternate Nexthop=C, Dist=4 • Keep shortest path only

  6. Protocols Examined (II): BGP D D:<A E F> D:<E F> D:<F> D: <B A E F> F B’s route to D: D:<E F> B A C E D:<C E F> • BGP is similar to DBF, but route includes entire path BGP: damping timer: 25 ~ 35 seconds BGP’: damping timer: 1~5 seconds † Route via A = <A E F> Route via C = <C E F>

  7. Outline for the rest of the talk • Introduction of RIP, DBF and BGP • Simulation results and lessons learned • Conclusion

  8. Simulation conducted 20 pkts/second • 7 by 7 mesh topologies similar those in [Baran64] • Simulated node degree range [3 ~ 16] • MeasurePacket loss, loops, path convergence time, throughput, and e2e delay.

  9. Packet Losses (I) : Observation • Packet losses of DBF, BGP’ and BGP decrease to zero at degree 6. • Richer connectivity helps RIP little. Packet Loss RIP DBF, BGP’ and BGP Node Degree

  10. Packet Loss(II): Lessons Learned RIP: D D DBF, BGP: F F B E C B E A C A • Keeping alternate paths • Connectivity Matters • no immediate available alternative due to poor connectivity and poison reverse • alternative is more likely with richer connectivity

  11. Packet Loss(III): Is an alternate path valid? D: < > D: < > D: < > D W C2 F V X U A B C E • Valid Alternate Paths: not using the failed link • Poison reverse and BGP’s path information are not enough! [Pei:Infocom2002] • Richer connectivity --> • reduces one single link’s impact • better availability of valid(but may be suboptimal) path

  12. Transient Loops(I): Observation • BGP has the most loops! • RIP has no loops • Richer connectivity reduces the chance of looping. BGP Losses due to loops BGP’ DBF Node Degree

  13. Transient Loops(II): Msg Propagation D: < > D: < > D: < > D: < > D: < > D D: <BC A E F> D: <B A E F> W F V U Y X D:<C B A E F> E C B A D:<C A E F> 30 seconds! • Damping timer slows the msg propagation, causing looping • Richer connectivity can reduce the chance of looping • More details in: “A Study of Transient Loops in BGP”

  14. Conclusion • Network’s Ultimate goal is to deliver happy packets, so Routing Protocols should • Maximize packet delivery during convergence • Achieve a good balance between packet delivery AND loop prevention, routing conv. time and routing overhead • Utilize the connectivity redundancy • Future work • Apply insights to BGP; study link state protocols, e2e TCP performance; Larger topologies, multiple pairs of S/D, multiple failures

  15. Questions?

  16. Instantaneous Throughput BGP’ Throughput(pkts/second BGP DBF RIP RIP Time

  17. Packet Delay During Convergence

  18. Forwarding Path Convergence time Time till the forwarding path from S to D stabilizes. Time till there is no routing msg. BGP:13 BGP:70 BGP’:2 BGP’:10 • BGP: no loss at degree 6 or higher • Shall we still tune MRAI timer to minimize convergence time(with the risk of increasing overhead)? Node Degree

More Related