1. Receiver-driven Layered Multicast Paper by- Steven McCanne, Van Jacobson and Martin Vetterli – ACM SIGCOMM 1996 Presented By – Manoj Sivakumar

2. Overview Introduction Approaches to Rate-Adaptive Multimedia Issues and challenges RLM - Details Performance Evaluation Conclusions

3. Introduction Consider a typical streaming Application What rate should the source send data at ? Internet Receiver Source 128 Kb/sX Kb/s

4. Approaches to Rate-Adaptive Multimedia Rate Adaptation at Source – based on available network capacity  Works well for a Unicast environment  How about multicast ? sourceReceiver 2 128 Kb/sX2 Kb/s Receiver 1 Receiver 3 X1 Kb/s X3 Kb/s

5. Example of Heterogeneity

6. Issues and Challenges Optimal link utilization Best possible service to all receivers Ability to cope with Congestion in the network All this should be done with just best effort service on the internet

7. Layered Approach Rather than sending a single encoded video signal the source sends several layers of encoded signal – each layer incrementally refining the quality of the signal Intermediate Routers drop higher layers when congestion occurs

8. Layered Approach Each layer is sent to one multicast group If a receiver wants higher quality – subscribes to all higher level layer multicast groups

9. Issue in Layered Approach No framework for explicit signaling between the receivers and routers A mechanism to adapt to both static heterogeneity and dynamic variations in network capacity is not present SSolution - RLM

10. RLM – Network Model Works with IP Multicast Assume  Best effort (packets may be out of order, lost or arbitrarily delayed)  Multicast (traffic flows only along links with downstream recipients)  Group oriented communication (senders do not know of receivers and receivers can come and go) Receivers may specify different senders

11. RLM - Video Streams One channel per layer Layers are additive Adding more channels gives better quality Adding more channels requires more bandwidth

12. RLM Sessions Each session composed of layers, with one layer per group Layers can be separate (i.e. each layer is higher quality) or additive (add all to get maximum quality)  Additive is more efficient

13. Router Mechanisms Dropping of packets  Drop less preferential packets first

14. RLM - Protocol Abstraction  on congestion, drop a layer  on spare capacity, add a layer

15. RLM – Adding and Dropping layers Drop layer when packet loss Add does not have counter-part signal Need to try adding at well-chosen times  Called join experiment

16. RLM – Adding and Dropping layers If join experiment fails  Drop layer, since causing congestion If join experiment succeeds  One step closer to operating level But join experiments can cause congestion  Only want to try when might succeed

17. RLM – Join Experiments Get lowest layer and start timer for next probe  Initially timer small  If higher level fails then increase timer duration else proceed to next layer and start time for the layer above it Repeat until optimum

18. RLM Join Experiment How to know is join experiment succeeded  Detection time

19. Detection Time Hard to estimate Can only be done experimentally Initially start with a large value Progressively update the detection time based on actual values

20. RLM - Issues with Joins Is this Scalable  What if each node does join experiments and the same time for different layers  Wrong info to node that requests lower layer if the other node had requested higher layer  Solution – Shared Learning

21. RLM – Shared Learning Each node broadcasts its intent to the group  Adv’s – other nodes can learn from the result of this node’s experiment  Reduction in simultaneous experiments Is this still foolproof ??

22. RLM - Evaluation Simulations performed in NS Video modeled as CBR Parameters  Bandwidth: 1.5 Mbps  Layers: 6, each 32 x 2m kbps (m = 0 … 5)  Queue management :Drop Tail  Queue Size (20 packets)  Packet size (1 Kbytes)  Latency (varies)  Topology (next slide)

23. RLM - Evaluation Topologies  1 – explore latency  2 – explore scalability  3 – heterogeneous with two sets  4 – large number of independent sessions

24. RLM – Performance Metrics Worse-case lost rate over varying time intervals  Short-term: how bad transient congestion is  Long-term: how often congestion occurs Throughput as percent of available  But will always be 100% eventually  So, look at time to reach optimal Note, neither alone is ok  Could have low loss, low throughput  High loss, high throughput Need to look at both

25. RLM – Performance Results Latency Results

26. RLM – Performance Results Latency Results

27. RLM – Performance Results Session Size

28. RLM – Performance Results Convergence rate

29. RLM – Performance Results Bandwidth Heterogeneity

30. Conclusions Possible Pitfalls  Shared Learning assumes only multicast traffic  Is this valid ??  Is congestion produced by Multicast traffic alone  Simulation does not other traffic requests!!

31. Conclusions Overall – a nice architecture and mechanism to regulate traffic and have the best utilization But still needs refinement

32. References S. McCanne, V. Jacobson, and M. Vetterli, "Receiver-driven layered multicast," in Proc. SIGCOMM'96, ACM, Stanford, CA, Aug. 1996, pp. 117--130.