The output of a switch, or, effective bandwidths for networks
Queueing Systems: Theory and Applications
Network decomposition: theory and practice
IEEE/ACM Transactions on Networking (TON)
| Hi-index | 0.00 |
We consider a two-stage queueing system where the first (upstream) queue serves many flows, of which a fixed set of flows arrive to the second (downstream) queue. We show that as the capacity and the number of flows aggregated at the upstream queue increases, the overflow probability at the downstream queue converges to that of a simplified single queue obtained by removing the upstream queue from the original two-stage queueing system. Earlier work shows such convergence for fluid traffic, by exploiting the large deviation result that the workload goes to zero almost surely, as the number of flows and capacity is scaled. However, the analysis is quite different and more difficult for the point process traffic considered in this paper. The reason is that for point process traffic the large deviation rate function need not be strictly positive (i.e., I(0)=0), hence the workload at the upstream queue may not go to zero even though the number of flows and capacity go to infinity. The results in this paper thus make it possible to decompose the original two-stage queueing system into a simple single-stage queueing system.