Coordinating Congestion Management and Bandwidth Sharing for Heterogeneous Data Streams

Many of the busiest servers in the Internet consist of clusters of nodes that serve out data of heterogeneous types. At a given point in time, a user could be receiving multiple concurrent data streams, both real-time and nonreal-time, that originate at a single server node, multiple nodes in the same server cluster, or nodes in separate clusters. In this paper, we argue that significant performance benefits can accrue from a coordinated approach to congestion management and bandwidth sharing across the concurrent data streams. We discuss several challenges that arise in this context and outline a coordination architecture that addresses these challenges.