1997 Volume 40 Issue 1 Pages 140-158
This paper evaluates cell-loss ratio in an output buffer of an ATM node based on the observed relative frequency of the number of cell arrivals during a fixed interval. The central issue of this evaluation problem is that the unique evaluation result of cell-loss ratio cannot be derived because the relative frequency of the number of cell arrivals does not completely describe the traffic characteristics of cell streams. Thus. this paper focuses on the derivation of the worst-case performance, that is, the cell-loss-ratio upper bound. Two cell-loss-ratio upper bounds are derived - one for when the cell streams are stationary, and one for when cell streams are stationary and ergodic. Numerical results show that cell streams with the same relative frequency of the number of cell arrivals can have quite different cell-loss ratios, and that the derived formula gives the actual upper bounds.