Study on the Effectiveness of the TCP Pacing Mechanism

The key idea proposed in this paper is to evaluate the effectiveness of TCP pacing mechanism minimizing the buffer requirements by routers. Based on references related to TFRC, buffer size needed in the bottleneck link, and analysis of the queue length in the router, the assumption that the router just needs less packets as buffer size, if the TCP traffic was paced than regularly TCP traffic in the bottleneck link shared by TCP and TFRC was made. Analysis of the queue length in the router was made. Simulation results show that the router can save at least 50% buffer size in the bottleneck link shared by pacing-based TCP and TFRC than in the bottleneck link shared by regular Reno TCP and TFRC, and when the utilization of the bottleneck reaches 90%, the loss rate hardly changes. The experiments are performed in both over-provisioned and under-provisioned network situations.