Minimizing consecutive packet loss in real-time ATM sessions

Abstract

Packet loss probability alone does not provide a complete characterization of the quality of a network's real-time packet sessions. Additional performance measures such as the average consecutive packet loss (the so called "average packet gap") must be considered. The authors propose to improve the packet gap performance of asynchronous transfer mode (ATM) sessions by minimizing the average packet gap present in the network's buffers. They introduce a simple packet dropping policy that, in steady state, minimizes: (i) the average packet gap, and (ii) the average packet buffering delay, when this delay can be reduced without increasing the average packet gap, or packet loss probability.