1 / 16

Multimedia

Multimedia. Robert Grimm New York University. Content: Multimedia Overview. Multimedia = audio and video Saroiu et al.—An Analysis of Internet Content Delivery Systems How is multimedia distributed over the Internet? How much is there? MacCanne et al.—Receiver-Driven Layered Multicast

alika-tyler
Télécharger la présentation

Multimedia

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. Multimedia Robert Grimm New York University

  2. Content: MultimediaOverview • Multimedia = audio and video • Saroiu et al.—An Analysis of Internet Content Delivery Systems • How is multimedia distributed over the Internet? • How much is there? • MacCanne et al.—Receiver-Driven Layered Multicast • How to best stream multimedia across the Internet?

  3. Stefan Saroiu’s OSDI Talk

  4. Based on broadcast model One server, many clients Clients subscribe to streams Basic problem: Network heterogeneity One approach: Fixed rate, least common denominator Streaming Multimedia

  5. Better Approach:Layered Transmission Scheme • Basic idea: Encode signal in many layers • Each layer provides better quality • Sum of layers represents a session • Cumulative layers • Independent layers • Simulcast

  6. Three assumptions Best effort, multipoint packet delivery Efficiency of IP Multicast Group-oriented communication Important issue: router drop policy Underlying Network Model

  7. Basic control loop On congestion, drop a layer On space capacity, add a layer The RLM Protocol

  8. Capacity Inference • One option: Monitor link utilization in network • Problem: requires changes to network • Another option: Actively probe network • Join-experiments • Issues with join-experiments • Adaptability • Scalability

  9. Goal Perform infrequently when likely to fail Perform frequently when likely to succeed Algorithm Join-timer for each layer Exponential backoff for problematic layers Join-Experiment Adaptability

  10. Join-Experiment Adaptability(cont.) • How to correlate join-experiment with outcome? • Need to chose appropriate detection-time • Unknown • Variable • Use estimator • Initialize conservatively • Adjust based on failed join-experiments

  11. Join-Experiment Scalability • Issue: interaction of independent join-experiments • Add congestion • Interfere with each other • Approach: scale frequency with group size • But, what about convergence?

  12. Receiver notifies group of join-experiment On congestion, other receivers increase corresponding join-timer Conservative Local Join-Experiment ScalabilityShared Learning

  13. Join-experiments are not completely exclusionary Lower or equal level experiments may overlap What about router drop policy? More on Shared Learning

  14. Based on simulations (ns) Two metrics Worst-case short-term loss rate Convergence time to sustainablethroughput Four topologies Latency scalability Session scalability Bandwidth heterogeneity Superposition Evaluation

  15. Results • RLM • Is sensitive to transmission latency • Scales with group size • Though, convergence time increases! • Supports bandwidth heterogeneity • Though, with increased loss rate • Supports simultaneous sessions • Though, allocation was often unfair

  16. What Did You Learn Today? • Content distribution in the Internet • Receiver-driven layered multicast

More Related