TCP agents and QoS measurements: Performance analysis on multi-source scenario

Abstract

In the present heterogeneous wired networks, reliability priority consideration is important. Quality of Service (QoS) varies according to congestion in the network and it can be avoided by different TCP Agents using congestion control algorithms. Previously, researches had only analyzed TCP agents, we extend their work to practical simulation for the multisource scenario to compare performance evaluation of different TCP agents: Tahoe, Reno, New-Reno, Sack1 and Vegas, based on QoS parameters like throughput, average end-to-end delay and packet loss by varying number of sources and Advertised Window (AW). Multisource environment is important from real world point of view. The selection of TCP Agents is based on type of application. We simulated TCP agents using ftp traffic and by increasing number of sources. We observed that Vegas works best for accurate delivery of data. Reno and Sack1 are suitable for real time streaming.