Abstract:
This paper is a follow up of a previous study done by [6], which examined the TCP performance over an OBS network model, using regenerative processes or the Markov renewal theory principles. A window of assembly times of about ±20% in reference to the optimal value, for the case of Wx < Wmax and a particular case of Wx = Wmax is set for an extended evaluation. This new study also considered the most common loss-based variants of the TCP protocol, such as Reno, New Reno and SACK. The evaluation results were contrasted by computer simulation, focused on TCP throughput and its corresponding theoretical and experimental error parameters. The goal of this study was to determine the degree of difference between simulated results and the respective analytical expressions obtained both in the previous article mentioned above and in the original study carried out by [3], due to the differences found in their mathematical expressions; considering the with and without delayed ACK scenarios for a comparative analysis between such variants.