My research is in computer network protocols and applications. I am specially intrigued by many decentralized algorithms adopted in Internet protocols, and the economic principles that govern their operations. A brief summary of some recent research results:

In the wireless area, there exists well establish models for single cell IEEE 802.11 networks (Bianchi et al), and theoretical models for wireless communication (Kumar et al). We had two recent publications (Y Gao first author). Our work establishes a new general fix-point model for the capacity of an IEEE802.11 multi-hop network. Our model takes into account of both contention between links used by one or more flows, as well as hidden node interference. This is a significant result for people studying IEEE 802.11-based multi-hop wireless networks. We validated our model using simulation, and applied our model to study different questions in multi-hop wireless networks. Another wireless network paper (first author RK Lam) analyzes how pricing can be applied when wireless nodes help others to gain connection to the Internet.

In the traffic measurement and analysis area, we have published some recent work that won a best student paper award at IEEE NOMS in 2006 (first author Y Hu). Network flow measurement is an important tool for Internet's management and traffic accounting. The current technology (netflow) does not scale and is vulnerable under traffic surge caused by various reasons (malicious traffic or flash crowd). Existing remedies can adapt to traffic intensity, but trades off accuracy for robustness. Our proposal is to find the best (based on information theoretic measure and practical considerations) flow aggregation to adapt to traffic intensity. The algorithm has been tested against real-life traffic traces to validate its merit. The algorithm and data structure are designed for real-time monitoring.

Internet is an open network accommodating a wide spectrum of applications. It is advocated (by IETF), however, that all traffic should use TCP-like congestion control (see literature on "TCP-friendly" congestion control). We question the wisdom of applying such a stringent approach to management congestion in the Internet. We have developed a framework for analyzing the co-existence of different traffic controls, and use it to demonstrate that it is advantageous to adopt some form of admission control for CBR-like applications (such as VoIP, or video conferencing). This is a somewhat controvertial topic, but we believe will become an important contribution in the long run (first author DM Chiu, or SW Tam).

We have published several papers in the area of P2P networks and overlay networks. Our major contributions include (a) a best student paper (in Performance 2005) that analyzes the optimality and interaction of multiple overlay networks in the Internet (WJ Jiang first author), and (b) a series of papers on modeling BitTorrent-like algorithms (B Fan first author). The BT-like algorithms have strong significance not only in content distribution, but also in new multi-path Internet protocol design. We also have some other publications in this topic area, including (c) an invited paper on applying network coding to P2P networking (DM Chiu first author); (d) the comparison of erasure coding versus whole file replication in networks with low availability peers (WK Lin first author), and (e) a recent paper on the problem of authenticating the content when network coding is applied to block-based content distribution protocols; we suggest an in-band authentication method that is significantly more efficient and secure than the source-based authentication, similar to that used in the original BT-like protocols.

We are also actively researching in the area of ISP traffic engineering and peering strategies (H Wang first author). We are one of the first to point out that multiple ISPs doing traffic balancing in multi-homing scenarios may cause oscillations and proposed guidelines to remedy the situation. Today's ISP peering is based on a transit service model, realized by policy-based inter-domain routing (BGP). The advent, and dominance of P2P traffic can significantly change the transit service model the ISPs have in mind. We developed traffic model and game-theoretic models to study how P2P traffic may affect ISP peering stability.

We are also interested in other network economic issues. We wrote a couple of papers (WY Ng first author) to study the viability of P2P networks based on a model of peers' common interests. In another paper (first author RK Lam), we analyzed how pricing can be applied when wireless nodes help others to gain connection to the Internet.