1 / 24

Kangaroo: Video Seeking in P2P Systems

Kangaroo: Video Seeking in P2P Systems. Xiaoyuan Yang † , Minas Gjoka ¶ , Parminder Chhabra † , Athina Markopoulou ¶ , Pablo Rodriguez †. ¶ University of California, Irvine. † Telefonica Research. Outline. Motivation and contributions Related work Architecture Peers Tracker

Roberta
Download Presentation

Kangaroo: Video Seeking in P2P Systems

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. Kangaroo: Video Seeking in P2P Systems Xiaoyuan Yang †, Minas Gjoka ¶, Parminder Chhabra †, Athina Markopoulou ¶, Pablo Rodriguez † ¶ University of California, Irvine † Telefonica Research

  2. Outline • Motivation and contributions • Related work • Architecture • Peers • Tracker • Experimental evaluation • Tracker • Segment Scheduler • Conclusion and future work

  3. Peer-to-peerChallenges • File sharing • reduces the service provider’s cost • allows scalability • Live Streaming • reduces upload cost, particularly important for • high quality videos, popular events • challenge: time sensitivity • Video on Demand • challenge: lack of synchronization between peers • Video on Demand with jump operations • enhances user experience • additional challenge: constant neighborhood re-adjustments

  4. Problem statement • Build a hybrid Video on Demand P2P system that • supports jumps • provides low buffering times • high swarming throughput • without overly provisioned peers and without aggressive prefetching

  5. Video on Demand P2P Related Work • Prior Work Bulletmedia - N. Vratonjic, P. Gupta, N. Knezevic, D. Kostic, A. Rowstron, “Enabling DVD-like features in P2P Video-on-Demand Systems”, in ACM P2P-TV Workshop 2007. Gridcast - B. Cheng, X. Liu, Z. Zhang, H. Jin, “A Measurement Study of a Peer-to-Peer Video-on-Demand System ”, in IPTPS 2007. PPLive – Y. Huang, T.Z. J. Fu, D.M Chiu, K.C.S. Lui, C. Huang, “Challenges, Design and Analysis of a Large-scale P2P VoD System”, in Sigcomm 2008

  6. ArchitectureKey Design Choices • Mesh-based P2P system with a pull model. • small segment size (64KB) • small active set of neighbors • Peers • adaptive hybrid segment scheduler • neighborhood manager • Smarttracker • smart neighbor selection • history based neighbor selection

  7. Architecture – Peers (1)Segment Scheduler • Segment scheduler decides what segment to download next • greedy strategy chooses segments for sequential playback • altruistic strategy chooses local-rarest segments • We propose a dynamic hybrid segment scheduler • starts with 80% greedy and 20% altruistic • ratio of sequential vs local-rarest segments varies dynamically depending on buffer size and segment deadline

  8. Architecture – Peers (2)Neighborhood manager + Peer Selection • Neighborhood manager • maintains a “healthy neighborhood’ • limits the number of active connection. • requests segments from “least useful” neighbors • Dynamic batching of “Have” messages • important reduction of control traffic

  9. Architecture – Tracker (1) • Observations: • The need of peers is guided by playback point • Peers play sequentially between jumps • Smart tracker meshes together peers that have content to exchange • Smart Neighbor Selection (SNS) returns list of peers at the same playback point • History Neighbor Selection (HNS) returns list of peers that contain needed segments

  10. Architecture – Tracker (2)Smart Neighborhood Selection (SNS) p=0 p=4 p=7 p=9 Peer A joins t=2, p=0 p=0 p=4 p=9 Peer B joins t=4, p=0 Play Point p 4 3 2 1 0 -1 -2 -3 -4 t=2 9 t=4 A B 7 4 3 2 1 0 -1 -2 -3 -4 4 t=6 A A B 1 4 3 2 1 0 -1 -2 -3 -4 4 6 8 2 Wall time t -2 A t=8 B -4 B

  11. Architecture – Tracker (3)History Neighborhood Selection (HNS) p=0 p=4,t=6 p=7,t=6 p=9,t=8 Peer A joins t=2 p=0 p=4,t=8 p=9,t=8 Peer B joins t=4 Wall time t=4 10 0 1 2 3 4 5 6 7 8 9 A A A B

  12. Architecture – Tracker (4)History Neighborhood Selection (HNS) p=0 p=4,t=6 p=7,t=6 p=9,t=8 Peer A joins t=2 p=0 p=4,t=8 p=9,t=8 Peer B joins t=4 Wall time t=7 10 0 1 2 3 4 5 6 7 8 9 A A A A A A B B B B

  13. Experimental evaluation • Experiment setup • peers rate limited at 1.5Mbps • playback rate of 1Mbps. • peer neighborhood size: 10-15 • 5 active download/upload connections • network emulated with a Modelnet cluster of 10 machines connected in a local Gigabit LAN • user behavior emulated from real traces collected from a live commercial IPTV service • Performance metrics • jump delay • seeder upload bandwidth

  14. Tracker performanceSimulations • 350 sessions from the most popular video • peer arrival a Poisson process with λ=1 peers/sec

  15. Tracker performanceSystem experiments • Comparisons with Random Neighbor Selection: • 3.5% reduction of segments uploaded at the seeder with SNS • 22% reduction of segments uploaded at the seeder with SNS+HNS

  16. Tracker Scalability • Tracker scalability depends on number of requests received by tracker: • every jump operation • triggered by neighborhood health evaluation • After experimental tests • we observed responses < 0.1ms for user behavior of 16,000 “dumb” peers • evaluated the trade-off between tracker response time and good topology connectivity

  17. Segment Scheduler performanceJump Delay • Greedy policy achieves the lowest delay • Adaptive hybrid allocates at least some bandwidth to download rare segments

  18. Segment scheduler performanceSeeder Load • Proposed adaptive hybrid best compromise between low seeder load and jump delay

  19. Conclusion • Designed a VoD system that provides good user experience without over-provisioning • Key mechanisms for this simple design • a smart scalable tracker • an adaptive hybrid segment scheduler • “least useful” peer selection • System tested with real users during 2008 Olympic games.

  20. Future work • Extend Kangaroo to support adaptive video quality • Plans to deploy Kangaroo as a Content Distribution Network.

  21. Questions?

  22. More questions?

  23. Tracker scalabilityTests

  24. Dynamic batching

More Related