Congestion control for best-effort service: why we need a new paradigm

Abstract

The current congestion control paradigm assumes that end users will use a single mandated algorithm. While the work done in this area has proven to be of great value, we need to recognize as a community that this paradigm is clearly inappropriate for future public networks. The reformulation of congestion control for best-effort service is discussed. We are not attempting to design specific new congestion control algorithms. Instead, we are merely trying to articulate the design principles. Many of these principles have been discussed before; however, with ATM currently designing a best-effort service under the name available bit rate (ABR) and the increasing commercialization of the Internet, these issues warrant revisiting. We outline the service model for best-effort service, describe the set of mechanisms available to implement this service model and contrast their various roles. The implications of our findings for future network design are discussed including some well-known examples of congestion control mechanisms.