Intelligent Grouping of TCP Flows for Coordinated Congestion Management
Hoi Chang, Rajeev Gopalakrishna, Venkatesh Prabhakar · Purdue e-Pubs (Purdue University System) · 2001
End-to-end congestion control mechanisms help utilize network resources more optimally because they inherently cooperate with the congestion control mechanisms inside the network.Recent research has shown that coordinating congestion control strategies between flows can improve the performance of the group of flows as a whole.An important aspect of such congestion management is identifying flows to be grouped together and the coordinated congestion technique to be adopted for the group.Currently Dows between the same end systems are grouped togther.Here, we attempt to provide an efficient, dynamic grouping mechanism for a set of flows originating from an end system on the basis of shared points of congestion.We use a correlation of the TCP round trip times (RTTs) to provide an estimate of the shared points of congestion.