Extended Evaluation of Loss-Based TCP Variants Performance Over Optical Burst Switching (OBS) Networks for Case W x =W max

This article aims to show an extended analysis of the previous study carried out by [5], where the performance of TCP on an OBS network model was addressed based on regenerative processes and Markov renewal theory (proposed in the literature) , and its evaluation by means of simulation using the network simulator program ns-2; However, in this new study, an expanded version of TCP throughput over OBS networks is provided for generic speed sources, taking into account an assembly time window of + -10%, in reference to the optimal assembly time for the particular case of Wx = Wmax, which in the previous study was left open for future work. As in [5], this new analysis is performed using the most common versions of the loss-based TCP protocol, such as Reno, New Reno and SACK, whose results are contrasted by means of computer simulation, in terms of the performance of TCP and its corresponding theoretical and experimental errors, bearing in mind the analytical expressions obtained both in the aforementioned previous article, as well as the original analysis carried out by [3], given the differences found in terms of its mathematical development; also maintaining the analysis for the scenarios with and without the delayed ACK option, for a comparative study between versions