Evaluation of TCP Based Congestion Control Algorithms Over High-Speed Networks

Abstract

The Additive Increase Multiplicative Decrease (AIMD) algorithm of the Transport Control Protocol (TCP) had worked remarkably well over low speed networks and had guaranteed fairness to the users all over these years, but at present, the demands for transferring large quantities of data over very high-speed networks are increasing at a tremendous rate. Because of its AIMD algorithm to control its window growth function accompanied by a slow response function which is inadequate over high-speed links, TCP has been proven to underutilize the available network bandwidth and leave a considerable amount of unused bandwidth. To overcome this limitation of TCP, the network research community came up with a number of TCP variants: HSTCP, STCP, BIC TCP, CUBIC, HTCP, and FAST TCP. All these protocols differ in the window growth policy to utilize the available bandwidth over a high-speed link. Various tests have shown that these protocols successfully utilize the link but at the same time they are not able to guarantee fairness to the other flows in the network. In this work, we aim to explore the following research questions: ○ Explore how tuning affects the performance of TCP and over 10G networks. ○ Compare TCP variants over a high-loss back-to-back environment In future, this work can be further extended in exploring the following two questions ○ Explore Performance Metrics for fair comparison of protocols over 10G back-to-back links ○ Move towards designing a congestion control protocol for back-to-back high-speed (Gigabit) links