Fast Approximate Analysis
of an ATM Multiplexer
Abstract
The performance of a statistical ATM multiplexer is analyzed with
respect to its buffer occupancy distribution.
The presumed traffic consists of classes of ON/OFF sources that are
distinguished by peak rate, activity, and burstiness.
The homogeneous (single-class) case is considered first, where
a fast, yet accurate method is developed for approximating the
steady-state occupancy probabilities.
Multiple classes are then addressed, maintaining an assumption
of homogeneity for peak rates only. By appropriately extending
the underlying model, a generalized version of the method
likewise provides accurate results.
Moreover, it remains computationally efficient if the number
of number of traffic classes is reasonable.
Compressed PostScript File:
itc95.ps.zip