1 / 31

Flow and Congestion Control for Reliable Multicast Communication In Wide-Area Networks

Flow and Congestion Control for Reliable Multicast Communication In Wide-Area Networks. A Doctoral Dissertation By Supratik Bhattacharyya. Talk Overview. General Problem Thesis Contributions Congestion Control for Single Multicast Group

treva
Download Presentation

Flow and Congestion Control for Reliable Multicast Communication In Wide-Area Networks

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. Flow and Congestion Control for Reliable Multicast CommunicationIn Wide-Area Networks A Doctoral Dissertation By Supratik Bhattacharyya

  2. Talk Overview • General Problem • Thesis Contributions • Congestion Control for Single Multicast Group • Efficient Flow Control Using Multiple Multicast Groups • Summary and Future Research Directions

  3. One-to-many reliable multicasting Transport-level techniques for congestion control flow control Focus Of Thesis Source Router R4 R1 R3 R2

  4. Challenges - many rcvrs, many network paths : Heterogeneity links, receiver capabilities Scale feedback implosion Fairness how to share bandwidth with unicast Multicast Flow/Congestion Control : a hard problem Source R1 R3 R2 R4 : end-to-end feedback

  5. Talk Overview • General Problem • Thesis Contributions • Congestion Control for Single Multicast Group • Efficient Flow Control Using Multiple Multicast Groups • Summary and Future Research Directions

  6. Thesis Contributions • Source-based Congestion Control : • identified and analyzed the Loss Path Multiplicity problem • identified a fair and scalable approach • formulated an axiomatic approach towards multicast congestion control • developed novel technique for responding to packet loss indications • designed a TCP-friendly protocol (NCA) for an active services architecture

  7. Thesis Contributions • Flow-control: • developed bulk data transfer approach using multiple multicast groups. • proposed and evaluated algorithms for determining transmission rate of each multicast group.

  8. Talk Overview • General Problem • Thesis Contributions • Congestion Control for Single Multicast Group • Efficient Flow Control Using Multiple Multicast Groups • Summary and Future Research Directions

  9. Challenge : How to aggregate feedback into single rate control decision Congestion signals (CS): filtered versions of loss indications (LI)  : congestion signal probability filters can be distributed Feedback Aggregation congestion signal (CS) loss indications (LI) rate change Rate control algorithm filter 

  10. Copies of same packet lost on many network paths Set of receivers treated as single aggregate receiver Example : n : no. of receivers p : loss prob. on link to each rcvr. : congestion signal probability LI LI R3 R1 Problem : Loss Path Multiplicity (LPM)  ?   1 as n  R2

  11. . . . How Severe is the LPM Problem? Example : end-to-end loss prob. = p=0.05 • Severe degradationin throughput with - • no. of receivers • independent losses f : fraction of end-to-end loss on independent link

  12. Feedback Aggregation/Filtering :Related Work • Restrict response to one LI per time interval T • Montgomery 1997 • Restrict response to subset of receivers : • choose K rcvrs out of N asrepresentatives • Delucia et al. 1997 • Reduce response to each LI : • Golestani, Bhattacharyya 1998, Delucia et al. 1997 Q :How much bandwidth should a multicast session get?

  13. Challenge : How to achieve “fair” sharing among multicast and unicast sessions Multicast allocation according to “worst” end-to-end path Multicast session shares equally with a unicast session on its “worst” end-to-end path. L2 L1 “Fair” Bandwidth Sharing Ucast1 Ucast2 Mcast L2 L1 - 1 Mbps, L2 - 2 Mbps

  14. Background : End-to-end Rate Control Algorithms : rate after i-th update • Additive increase, multiplicative decrease : on congestion signal : else, per T : • We derive average session throughput B

  15. Solution to LPM Problem : Our Approach • Worst Estimate-based Tracking (WET) : • Identify (estimate) most congested/ ”worst” receiver • Respond to LIs from only “worst” receiver • Simulations show that WET • prevents throttling of multicast transmission rate • allows fair bandwidth sharing

  16. WET is one way of designing a Loss Indication Filter (LIF) Qn : Given our fairness goal, can we formulate general rules for LIF design? Architecture for Loss Indication-based Multicast Congestion Control loss indications (LI) rate change congestion signal (CS) Rate control algorithm filter 

  17. N receivers, loss probabilities = unicast bandwidth on path to rcvr i Axiom 1 :IfN=1, then = Axiom 2 : If then Axiom 3 : As Goal : Multicast bandwidth allocation must be worst-path fair Axiomatic Approach for Loss Indication Filter Design . . . 2 1 N

  18. Linear Proportional Response (LPR) • Receiver i periodically reports loss count over W packets ( estimates ) • On LI from receiver i, source reduces rate with probability • Showed that LPR satisfies all three axioms

  19. Related : Random Listening Algorithm (RLA) [Wang98] Analytic Result : LPR provides tighter upper bound on r LPR : RLA : Comparison of LPR and RLA

  20. Summary of Results • LPR “more fair” than RLA for realistic W (~100 packets) • Steady State : • WET is closest to fairness goal • LPR is close to WET • RLA can be extremely unfair • Transient Behavior : • LPR, RLA respond faster to changes in network conditions than WET

  21. Transient Behavior 5 mcast over all links • At t=300 sec, two multicast sessions stop receiving feedback from receivers at the end of L1 L10 L1 L2 10 ucast 5 ucast 5 ucast . . . Loss probability on Link L2

  22. Talk Overview • General Problem • Thesis Contributions • Congestion Control for Single Multicast Group • Efficient Flow Control Using Multiple Multicast Groups • Summary and Future Research Directions

  23. Challenge : reliable delivery of finite volume of data diverse receive-rates Goal : minimize average completiontime Approach : multiple IP multicast groups (channels) Flow-controlled Bulk Data Transfer :Overview R3=3 R1=1 R2=2 R4=4

  24. Q : How to : assign channel rates? assign receivers to channels? partition data among channels? Assumptions : error-free channels known, static receive-rate constraints Solution with unlimited channels : minimizes average completion time minimizes bandwidth Flow-controlled Bulk Data Transfer 2 pkts/sec 4 pkts/sec 1 pkt/sec R2 a R4 R1 b a a b c d R1,R2,R4 r1 = 1 a b d c r2 = 1 d b R2,R4 c d r3 = 2 R4

  25. Q : How to : assign channel rates? assign receivers to channels? partition data among channels? Assumptions : error-free channels known, static receive-rate constraints Solution with unlimited channels : minimizes average completion time minimizes bandwidth c d Flow-controlled Bulk Data Transfer 2 pkts/sec 4 pkts/sec 1 pkt/sec R2 a R4 R1 b a a c b c d R1,R2,R4 r1 = 1 a b d c r2 = 1 d b R2,R4 c d r3 = 2 R4

  26. Q : How to : assign channel rates? assign receivers to channels? partition data among channels? Assumptions : error-free channels known, static receive-rate constraints Solution withunlimited channels : minimizes average completion time minimizes bandwidth c d Flow-controlled Bulk Data Transfer 2 pkts/sec 4 pkts/sec 1 pkt/sec R2 a R4 R1 b a b a c b d c d R1,R2,R4 r1 = 1 a b d c r2 = 1 d b R2,R4 c d r3 = 2 R4

  27. Summary of Results • Developed solution for minimizing average completion time with N receivers and K channels • Developed simple rate assignment algorithms that • scale well to large number of receivers • have close to optimal average completion time • make efficient use of network bandwidth • Showed that small number of multicast groups sufficient for above algorithms

  28. Summary of Contributions • Source-based Congestion Control : • identified and analyzed the Loss Path Multiplicity problem • identified a fair and scalable approach • formulated an axiomatic approach towards multicast congestion control • developed novel technique for responding to packet loss indications • designed a TCP-friendly protocol (NCA) for an active services architecture

  29. Summary of Contributions • Flow-control: • developed bulk data transfer approach using multiple multicast groups. • proposed and evaluated algorithms for determining transmission rate of each multicast group.

  30. Future Research Directions : Congestion Control • WET : • How can the source detect changes in network congestion levels in a timely fashion? • LPR : • Can steady state performance be improved? • Can the NCA protocol be based on LPR instead of WET? • NCA : • implementation details - start-up, nominee changeover, etc.

  31. Future Research Directions :Flow Control • Flow-controlled bulk data transfer : • evaluate performance when sender has imperfect knowledge of receive-rates • explore feasibility of our approach in a practical setting • Synergy with per-group congestion control techniques

More Related