In: Performance Evaluation Review, Vol. 26, No. 2, pages 38-46. 1998.
Abstract: This paper presents a performance evaluation of windowing mechanisms in world-wide web applications. Previously, such mechanisms have been studied be means of measurements only, however, given suitable tool support, it is shown that such evaluations can also be performed using infinite-state stochastic Petri nets. The paper briefly presents this class of stochastic Petri nets as well the approach for solving the underlying infinite-state Markov chain using matrix-geometric methods. Then, a model of the TCP slow-start congestion avoidance mechanism is presented, subject to a (recently published) typical world-wide web workload. The model is parameterized using measurement data for a national connection and an overseas connection. The study shows how the maximum congestion window size, the connection release timeout and the packet loss probability influence the expected number of buffered segments at the server, the connection setup rate and the connection time.
Keywords: congestion control, matrix-geometric method, stochastic Petri nets, window flow control.