1. Proxy-based Distribution of Streaming Video over Unicast/Multicast Connections B. Wang, S. Sen, M. Adler and D. Towsley University of Massachusetts Presented by Nera Liu

2. Agenda Introduction Problem Setting and Model Optimal Proxy Cache Allocation Proxy-Assisted Transmission Schemes Performance Evaluation Conclusion

3. Introduction With the emergence of broadband network, streaming video over the Internet becomes feasible nowadays. However, due to its high bandwidth and large client requirement, it is inefficient in traditional VoD system i.e. unicast-based system.

4. Introduction In designing scalable and cost effective VoD system, researchers mainly focused on two different approaches Open-loop architectures System performance is independent of system load It achieves better performance under heavy system load Closed-loop architectures System performance is dependent of system load It achieves better performance under light system load.

5. Introduction Assumption of previous works It is designed based on a multicast-enabled network. However, IP multicast deployment in the Internet is slow.

6. Introduction The use of proxy cache is an orthogonal technique for reducing This technique is proved to be efficient for delivering Web objects. Similar works have been proposed in which a certain portion of video file is stored in either client side and proxy server

7. Introduction S. Sen, J. Rexford, and D. Towsley, “Proxy prefix caching for multimedia streams,” in Proc. IEEE INFOCOM, Apr 1999. J. Almedia, D. Eager, and M. Vernon, “A hybrid caching strategy for streaming media files,” in Proc. SPIE/ACM Conference on Multimedia Computing and Networking, Jan 2001 Y. Wang, Z.-L. Zhang, D. Du and D. Su, “A network conscious approach to end-to-end video delivery over wide area networks using proxy servers,” in Proc. IEEE INFOCOM, Apr 1998.

8. Introduction What are suitable proxy-assisted closed-loop transmission schedules? For a given transmission schedules, what is the optimal proxy prefix caching scheme that minimizes the transmission cost? What are the resources tradeoffs (proxy cache space and transmission bandwidth etc.) for the different transmission schedules?

9. Problem Setting The server-proxy path is unicast- enabled. The proxy-client path is multicast-enabled.

10. Problem Setting N CBR-encode video with different popularity f i Video i has a video bit rate b i, and length L i seconds and size n i unit

11. Design Goal Ci(vi) is the transmission cost per unit time for video i when a prefix of length vi of video i is cached in the proxy. Goal: Minimize the value of Ci(vi).

12. Optimal Proxy Cache Allocation Determine the storage vector (v1, v2 ….) that minimizes the aggregate transmission cost for a given proxy- assisted transmission scheme. In this paper, it proposes two different schemes Optimal 0-1 caching Optimal prefix caching

13. Optimal Proxy Cache Allocation Optimal 0-1 caching Video is either store entirely or not Saving = Ci(0) – Ci(L) Optimal prefix caching Only a prefix of video is stored in the proxy Saving (mi) = Ci(0) – Ci(mi), where mi represents a vector of possible prefix of video i storing in the proxy.

14. Optimal Proxy Cache Allocation The optimal problem where S is the storage size of the proxy

15. Proxy-Assisted Transmission Scheme The next step is how to determine Ci(vi) ? Unicast suffix batching (SBatch) Unicast patching with prefix caching (UPatch) Multicast patching with prefix caching (MPatch) Multicast merging with prefix caching (MMerge)

16. Unicast patching with prefix caching (UPatch)

17. Multicast patching with prefix caching (MPatch)

18. Performance Evaluation

21. Optimal caching > Priority Prefix caching > Optimal 0-1 caching MMerge > MPatch > UPatch > SBatch

22. Conclusion Under optimal proxy caching, even a small cache is sufficient to realize the transmission cost saving. The selection of closed-loop transmission scheme coupled with optimal proxy caching can produce significant cost saving.