Performance characteristics of end-stations in an ATM network as viewed by applications and network

Abstract

Traffic shaping at a (traffic) source station is a prominent congestion control solution in ATM networks. The `leaky bucket' scheme in conjunction with a cell spacer is an extensively studied and used mechanism for such a traffic shaping. By dimensioning the leaky bucket and spacer parameters, a source station attempts to regulate the behavior of its own traffic in order to conform to a `good behavior' contract with the network. In this paper, we study the effect of the a single source's traffic characteristics and its leaky bucket parameters on the maximum burst size and the burst size distribution as seen by the network. The theoretical upper bound for the burst size is derived and simulation is used to study the effect of leaky bucket parameters on probability distribution of burst sizes. It is shown that while traffic shaping is done to abide to negotiated traffic characteristics, traffic shaping can lead to large queues of cells and result in large cell delay and delay jitter in the source station.