NETCOM Research Group

Multimedia content delivery is projected to be the biggest bandwidth consumer of the future Internet. For many years, the mechanism for content delivery envisioned by the networking community is network multicast.

Multicasting emulates traditional TV broadcasting; it is designed to be network efficient. But it falls short in at least two aspects: (a) it does not maximize throughput for content delivery; (b) like TV broadcasting, it does not provide on-demand access (i.e. Video-on-Demand).

A competing technology is CDN-assisted client-server VoD. While this gives us (b), it has serious scalability problems. In recent years, Peer-to-peer (P2P) content delivery has proven to be quite successful in achieving both (a) and (b). P2P networks have evolved from mostly "content sharing" platforms to become "content delivery" platforms, using interesting distributed algorithms to help achieve good trade-off between scalability, qualtiy of service and economical resource allocation. In modern P2P platforms, the resources managed by the P2P system is not only the peer bandwidth, but also storage (for content replication) and CPU (for network coding).

In this seminar, I will give an introduction to this active research area, and explain some of our research contributions. I will explain the concept of the "streaming capacity" of a P2P network, originally studied in the thesis of Mundinger; I will then discuss in detail a number of classic papers modeling and analyzing P2P algorithms that try to achieve this capacity, and make the trade-off between system throughput and fairness.

These include the classic work by authors including Srikant, Massoulie and Hajek. For our own work, we will discuss work by B Fan et al, and YP Zhou et al.

I will also describe my recent collaboration with other research labs in building a large scale simulation tool for evaluating a city-wide P2P content distribution platform.

To close the seminar sessions, I will also give some additional talks on other related research interests. One topic will be the analysis of academic publication records for papers in the networking area in the past 20 years.

Another topic will be on some recent work in network economics.

Reference material and recommended reading

Who is Prof. Dr. Dah Ming Chiu?

Professor Dah Ming Chiu received his undergraduate degree from Imperial College, London, and his PhD degree from Harvard University.

Subsequently, he worked for AT&T Bell Labs, Digital Equipment Corporation, and Sun Labs. He joined the Chinese University of Hong Kong in 2002.

Dah Ming is awarded an IEEE Fellow, recognizing his work in studying network resource allocation problems as in congestion control starting in the 1980s. His current research interest is in P2P networks, network economics, and Internet engineering (including measurement, analysis and future Internet architecture).

He is an editor for IEEE/ACM Transaction for Networking, and has served on the TPC for many top networking conferences in recent years. He is currently also serving as the chairman of his department.

More details at: http://home.ie.cuhk.edu.hk/~dmchiu/

The conference will be conducted in English