Comparative study of TCP congestion control algorithms

TCP (Transmission Control Protocol), is a reliable transport protocol for the transport layer of TCP / IP model, it is the most used transport protocol since it implements panoply of mechanisms ensuring good data transfer.Nowadays, the internet knows a huge growth and therefore it becomes more and more difficult to guarantee the continuity of the services to a very large number of users.This is why a lot of research has been carried out in order to improve the functioning of TCP generally and congestion control more precisely.Several congestion control mechanisms have been proposed to improve the performance of TCP.But, it still suffers from unsatisfactory performances.This is why we will try in this article to conduct a study -based on the analysis of some metrics, packet drop, latency and throughput -for the analysis and comparison of the most powerful of the algorithms proposed in this sense (Tahoe, Reno, New Reno, Vegas, Sack, Fack), in order to identify their advantages and limits under congested environments.So, for the simulation we opted for the NS2 simulator by applying 24 different scenarios for each algorithm.The study showed that TCP Tahoe, Reno, New Reno and Sack are loss-based; they favor the loss of packets to guarantee a short latency.While TCP Vegas is delay-based; it's recommended for applications that require reliable packet transfer.The results of this study will form the basis of future work on the development of a robust algorithm which combines both the advantages of the studied algorithms, and which has the power to share bandwidth fairly with aggressive algorithms such as TCP Reno in order to ensure a good congestion control.