DELAY PERFORMANCE OF SOME SCHEDULING STRATEGIES IN AN INPUT QUEUING ATM SWITCH WITH MULTICLASS BURSTY TRAFFIC

We consider an N x N nonblocking, space division, input queuing asynch ronous transfer mode (ATM) cell switch, and a class of Markovian model s for cell arrivals on each of its inputs, The traffic at each input c omprises geometrically distributed bursts of cells, each burst destine d for a particular output, The inputs differ in the burstiness of the offered traffic, with burstiness being characterized in terms of the a verage burst length, We analyze burst delays in the situation in which some inputs receive traffic with low burstiness and others receive tr affic with higher burstiness. Three policies for head-of-the-line cont ention resolution are studied: two static priority policies [viz,, sho rter-expected-burst-length-first (SEBF), longer-expected-burst-length- first (LEBF)] and random selection (RS), Direct queuing analysis is us ed to obtain approximations for asymptotic (as N --> infinity) high an d low priority mean burst delays with the priority policies, Simulatio n is used for obtaining mean burst delays for finite N and for the ran dom selection policy, Numerical results show that, as the traffic burs tiness increases, the asymptotic analysis can serve as a good approxim ation only for large switch sizes, Qualitative performance comparisons based on the asymptotic analysis are, however, found to continue to h old for finite switch sizes. It is found that the SEBF policy yields t he best delay performance over a wide range of loads, while RS lies in between, SEBF drastically reduces the delay of the less bursty traffi c (e,g,, distributed computing traffic) while only slightly increasing the delay of the more bursty traffic, e,g,, variable bit rate (VER) v ideo, LEBF causes severe degradation in the delay of less bursty traff ic, while only marginally improving the delays of the more bursty traf fic, RS can be an adequate compromise if there is no prior knowledge o f input traffic burstiness.