PERFORMANCE COMPARISON OF DIFFERENT TCP VARIANTS IN IP AND MPLS NETWOTKS

Abstract

This project presents an analysis of the Transmission Control Protocol (TCP) and its different variants on IP and MPLS (Multi-protocol Label Switching) networks. TCP provides a trustworthy end-to-end data transfer under changeable wired networks. To overcome the problem of unreliability of IP network, TCP is used. Many service providers are now moving to MPLS over internet to transfer data, preferring it over traditional transferring strategies. Different variants of TCP show varying behavior in best effort Internet Protocol networks. This research presents an extensive investigational study of TCP variants under IP and MPLS networks by focusing Tahoe, Reno, New Reno, Sack and Vegas under File Transfer Protocol (FTP) and Constant Bit Rate (CBR) traffics. For analytical results of the proposed solution; demonstration of the IP and MPLS network is simulated over a limited number of nodes. The flow of descriptors to maintain network topology determines average delay, variance of delay, packets sent, packets received, packet dropped, throughput and congestion window size. Conclusions are drawn based on the simulation results, while comparisons between them have been elaborated. Under different traffic flows, it has been found that all variants are unable to adapt MPLS features. However, Vegas has shown potential outcomes with almost invariable en-to-end delay after a brief time period of delay oscillation in the early phase of application. The unwavering end-to-end delay of Vegas under MPLS makes it a striking option for average to huge size real-time networks. This research work has identified open problems related to the TCP protocol and its variants. To limit the scope of the research portion, error rate has been introduced only in single flow of FTP. This analysis has helped in investigation of congestion window patterns of the TCP variants with respect to different error rates induced over the network.