My principal research interests have been in the areas of reliable multicast and congestion control, both at the transport level, and because of the overwhelming deployment challenges, even more so at the content distribution network/application layer multicast/messaging middleware level.  My particular passion has been for research that is driven by real world requirements, and has the potential to make a direct positive impact on the industry.

I was actively involved in the IETF standards body for a number of years, in the Reliable Multicast Transport working group.  This work was the result of the commercial experiences I had, combined with my reliable multicast research. 

For years, the primary research hurdle to driving standards for reliable multicast was solving the problem of multicast congestion control.  In helping attack this problem, we found that there was surprisingly little theoretical foundation underneath Internet congestion control, and we helped contribute to building a new theoretical foundation for congestion control.  My dissertation was on tree based reliable multicast and multicast congestion control.

As part of the team at GlobalCast, we designed and shipped the first commercial application layer multicast system in 1996.  I continued this work at Talarian, including helping design a next generation architecture for the core of our SmartSockets publish/subscribe product.  However, while both of these projects were on the leading edge of both academic research and commercial development, both projects were kept proprietary, so unfortunately there are no publicly available documents on either.  

Other research I have worked on over the years includes quantifying the "packet starvation effect" of Ethernet networks, and proposing a solution that both solves this problem and provides latency guarantees for real time traffic.  My original work in reliable multicast was also tied to distributed systems.  The Reliable Multicast Protocol (RMP) designed in 1991-1993, was one of the first reliable multicast protocols, and provided a wide range of distributed systems guarantees, such as total ordering and atomic delivery, with extremely high performance.  As part of my M.S. thesis, I also built one of the first distributed publish/subscribe systems (the Message Bus II) on top of RMP, in 1992.

One other field of research I was working in as I left high tech was in the area of "contract based object routing", which was looking at extending SOAP to provide a super-scaleable interface to publish/subscribe and other types of application layer multicast/overlay networking infrastructure, while solving the key problems of existing API standards such as Java Messaging Service (JMS)--namely, dependency on a single language and inability to bridge between the infrastructure of multiple software providers.

Copyright © 2007 Brian Whetten