Title:Worst-Case Services and State-Based Scheduling

Abstract:In this paper, we shed new light on a classical problem: given a slot-timed, constant-capacity server, to provide long-run service guarantees to competing flows of unit-sized tasks, what short-run scheduling decisions must be made? We model each flow's long-run guarantee as a worst-case service that maps each queued arrival vector recording the flow's cumulative task arrivals, including those initially queued, to a worst-case acceptable departure vector lower-bounding its cumulative task departures. We show these services to be states that can be updated as tasks arrive and depart, and introduce state-based scheduling. We find the schedulability condition that must be preserved to maintain all flows' long-run guarantees, and then use this condition to identify, in each slot, all short-run scheduling decisions that preserve schedulability.
This framework is general but computationally complex. To reduce its complexity, we consider three specializations. On the one hand, we show that when satisfactory short-run scheduling decisions exist, some special ones can be efficiently identified by maximizing the server's capacity slack. On the other hand, we show that a special class of worst-case services, min-plus services, can be efficiently specified and updated using properties of the min-plus algebra, and that this efficiency can be further improved to verge on practical viability by restricting attention to a further specialization, dual-curve services, which turn out to be dynamic extensions of service curves.