1. Video Multicast over the Internet Presented by: Liang-Yuh Wu Lung-Yuan Wu Hao-Hsiang Ku 12 / 6 / 2001 Bell Lab. And Georgia Institute of Technologies IEEE Network · March/April 1999

2. Agenda 1. Abstract 2. Basic idea 3. Single-stream video multicast 4. Replicated-stream video multicast 5. Layered video multicast 6. Replicated vs. Layered 7. Error Control 8. Concluding Remarks

3. Abstract  Multicast of video / future network service Video Conference Distance Learning Remote Presentation Media-on-Demand ( MOD,VOD )  Lacks support for QoS assurance  Heterogeneity of the Internet transmission Why?  Problem of “ Real time & Bandwidth ”

4. Base requires for Multicast video :  Mechanisms for Multicast data delivery  Ability to accommodate Real-time video Focus on:  1.How real-time video can be accommodated over Internet 2.Fairness Focus & Discussion ??

5. Some general approaches and issue (Real-time): ---Delay jitter, Data loss 1. Using QoS reserve resources 2. Using adaptive rate control “ Feedback mechanism ” 3.Buffering 4.Change compression parameters 5.Error control technology

6. Video Bit-Rate adaptation Adaptive Bit-Rate Video Multicast Options 1. Single Stream Adaptive Approach 2. Replicated Adaptive Streams Approach 3. Layered Video Streams Approach Basic idea

7. Single-stream Video Multicast The IVS Approach (in the H.261 encoder) Feedback information is based on packet loss measured at the receiver Packet loss is detected using RTP (Real-time Transport Control) RTP sends “reception reports” that provide feedback information

8. Single-stream Video Multicast(cont.) Feedback implosion problem Probabilistic multicast technique Probing mechanism Can’t provide fair treatment to multiple receiver in a heterogeneous environment Tradeoff between multicast and unicast

10. Replicated-Stream Video Multicast The DSG (Destination Set Grouping) Protocol The goal of the DSG: To improve fairness over a single-group feedback-controlled video multicast To address “scalability” By transmitting video of differing quality and differing data rate on different multicast channels and allowing receivers to select the most appropriate one.

12. Replicated-Stream Video Multicast (cont.) The DSG protocol has two main Components: Intra-stream protocol Inter-stream change protocol A DSG experiment result : Fairness among receiver is improved significantly over a single-group approach while incurring only a small additional bandwidth overhead.

14. Layered Video Multicast Video layering can be supported by many video compression techniques EX: MPEG-2 supports layered encoding by defining four scalable modes Layered Video Multicast Protocol Receiver-driven Layered Multicast (RLM) Hierarchical Rate control (HRC)

16. Receiver-driven Layered Multicast Receiver-based control Advantage: burden of adaptation is moved from the sender to the receiver Join-experiment result is successful result is failed

17. Receiver-driven Layered Multicast (cont.) The key to scalability in layered multicast is “adding or dropping a layer” Shared Learning Advantage: Disadvantage: unnecessary bandwidth and message processing overhead Too much state information

18. Hierarchical Rate control (HRC) Layered Video Multicast with retransmissions (LVMR) two key contributions: retransmitting lost packets adapting to network congestion and heterogeneity using HRC Hierarchical Rate control (HRC) to distribute the information between the sender, receivers, and some agents each entity maintains only the information relevant to itself

19. Hierarchical Rate control (HRC) (cont.) Comparing with RLM: allow receiver to maintain minimal state information decrease control traffic on the multicast session multiple experiments to be conducted simultaneously drop the correct layer(s) during congestion in most case

20. Hierarchical Rate control (HRC) (cont.) In addition to avoid the above drawback of RLM Comprehensive group knowledge base Collaborative layer drop decrease layer oscillation achieves more effective rate adaptation maintains better video reception quality Add-layer experiment decrease unnecessary add-layer experiment provide smoother video quality

21. Replicated-stream VS. Layered video For video multicast, layering is determined more by the coding requirements than by the bandwidth requirements Replicated-stream VS Layered video: bandwidth economy Processing overhead

22. Error Control 1.Layered Video Multicast with Retransmission (LVMR) 2.Structure-Oriented Resilient Multicast (STORM) 3.Client-Server Architecture

23. The Key Idea In LVMR Use a statically configured logical tree Set Designated Receivers (DRs) at each level Improve efficiency by no asking for retransmission Improve response time by sending immediate message and multicast retransmissions Use Buffers Combine retransmission mechanisms

26. The Key Idea In STROM Each receiver to dynamically select the best possible DR Use a dynamically logical tree Receiver decide the latency or reliability

27. The Steps of STORM 1.Build the recovery structure 2.Selection of Parent Nodes 3.Adapting the structure

28. The Key Idea In Client-Server Architecture Separate the actual senders and receivers. Receiver either get the original or repaired video stream. The retransmit server and the repair server can be organized into LVMR Using RTP Repair Buffer

30. Concluding Remarks Network congestion lead to degrade the video quality. Non-adaptive streams of video data don not share resource well in a best-effort network.

31. Future Work To build a mechanisms which provide for the co-existence and resource sharing of video multicast stream. Formalization of the notion of fairness Understanding the effect of pricing on the behavior of receiver in a multicast video environment.

32. The End Thank you so much!!