Paper review: TCP Vegas: End to End Congestion Avoidance on a Global Internet

Reviewer: Jason Oh

  1. This paper makes an attempt of introducing improvements to the existing TCP protocol, focusing primarily on the Reno implementation.
  2. The main contribution of this paper is to propose a new algorithm to modify and improve the TCP protocol. Although not particularly new, this area is important because the growing number of users and applications of the Internet make it increasingly important to improve its performance. This problem is and will continue to be important for as long as people find new and more bandwidth-intensive ways to use the Internet. Brakmo and Peterson come up with their improvements by examining where the Reno implementation falls short, and they consider new approaches to the problems that Reno solved.
  3. The main idea of this paper is that the Vegas implementation is a (nearly) unqualified success, superior in (nearly) every way to the Reno implementation.
  4. This paper rates around a 4 in terms of contribution because the performance improvements are significant, although Brakmo and Peterson are less than completely forthcoming with the data that they collected.
  5. Their methodology is rather convincing, since they report considerable performance improvements over the Reno implementation. The authors provide analyses to back up their algorithm, and their conclusions follow logically from their data.