Causal Scheduling of Multiclass Traffic With Deadlines and Priorities.
ILLINOIS UNIV AT URBANA COORDINATED SCIENCE LAB
Pagination or Media Count:
This thesis analyzes causal scheduling and scheduling-dropping policies in a discrete time model. Packets from different priority classes arrive with arbitrary deadlines. The packets must be scheduled before their deadlines. We present three sets of results. The first set is divided into two parts. We first characterize all causal scheduling policies which maximize the throughput whatever the sequence of packets arriving to be scheduled. We then extend the analysis to causal scheduling-dropping policies, that is, scheduling policies which can drop packets before there expiration without scheduling them. We characterize all such scheduling policies that maximize the throughput and those that do so while minimizing the buffer occupancy. The second set of results considers a two class model. We characterize all causal scheduling and scheduling-dropping policies that maximize the high-priority class throughput subject to maximizing the overall throughput For the scheduling dropping policies, we characterize those policies that minimize the buffer occupancy. We finally analyze the more general multi-class case, characterize all causal scheduling policies that optimize two novel optimality criteria, and present two simple scheduling-dropping policies that optimize those criteria.
- Administration and Management