The WING Blog

The Web and Internetworking Group at BU/CS

May

15

What are your favorite 1999-2001 Sigmetrics papers?

By Azer Bestavros

Here are mine — comments?

Y. Chu, S.G. Rao, and H. Zhang. A Case for End System Multicast. In Proceedings of ACM Sigmetrics, June 2000.
http://academic.research.microsoft.com/Publication/247484/a-case-for-end-system-multicast

Lixin Gao and Jennifer Rexford. Stable Internet routing without global coordination. In Proceedings of ACM Sigmetrics, June 2000.
http://academic.research.microsoft.com/Publication/234493/stable-internet-routing-without-global-coordination

Thomas Bonald and Laurent Massoulié. Impact of fairness on Internet performance. In Proceedings of ACM Sigmetrics, June 2001.
http://academic.research.microsoft.com/Publication/680569/impact-of-fairness-on-internet-performance

Steven H. Low, Larry Peterson, and Limin Wang. Understanding TCP vegas: a duality model. In Proceedings of ACM Sigmetrics, June 2001.
http://academic.research.microsoft.com/Publication/115673/understanding-tcp-vegas-a-duality-model

3 Responses so far

These are all very good papers who have made an impact.

End-system multicast came along after a decade of effort on IP multicast — trying to build multicast into the routing protocols of the Internet. It eventually became clear that the complexity of the joint economic/technical issues were just too great to permit IP multicast to take root. Then along comes ESM, which made the problem so simple: just accept that there will be a bit of inefficiency compared to in-network support, and voila! It all works pretty well, and deploying is easy. I see this realization as prescient of the rise of Napster and its academic offspring, P2P.

The Gao-Rexford (gotta just call it that now) paper established a sufficient condition for BGP stability. ’nuff said. It mattered then and it still matters now.

The TCP Vegas paper illustrated the idea that a flow control algorithm might actually be solving an optimization problem, and that we can understand flow control by casting it as an optimization. This idea had been put forward by Kelly in the same timeframe, but the TCP Vegas paper really “revived” Vegas (it had been previously sort of shouted-down by the community) using this powerful analysis approach.

The ESM paper came out at the “right time”. Application-layer multicast was “in the air” at the time — e.g., . Paul Francis’ Yoid system in 1999. The reason ESM is cited so much is because of the many subsequent pieces of work that performed functionality that had been previously found below the application layer (multicast, transport) in the application layer, but I would not attribute the paper as being the seed/cause for all of the following work…

It’s a funny thing why some papers end up being referenced (or over referenced 😉

Agree – all excellent papers and work. Few thoughts:

ESM is an example of communicating processes forming their own network and collaborating to provide some service (here, multicast streaming) using communication services of the underlying Internet. You can see some recursion 😉 Our RINA architecture tries to formalize this recursive service-based structure…

In what is emerging as service-based architectures, it doesn’t seem BGP will be needed since what a “layer” provides is a “service”, not a route… And transport becomes a composition of services so we should be able to do better than relying on “implicit” feedback…

It’s amazing how solutions change (and often become simpler) if we got the right architecture…

Leave a comment