Heuristic optimization of scheduling and allocation for distributed systems with soft deadlines

Abstract

This paper studies optimal deployment and priorities for a class of distributed real-time systems which have complex server tasks, many concurrent scenarios, operations with deterministic or stochastic execution demands, arbitrary precedence between operations, and hard or soft deadline requirements. The soft deadlines take the form of a required percentage of responses falling within the deadline. This work improves on an earlier optimization approach which was only applied to hard deadlines. As before, heuristic measures derived from solutions of layered queueing models are used to guide step-by-step improvement of the priorities and allocation, searching for a feasible solution which meets the soft deadline requirements. Effectiveness is demonstrated on a range of examples including thousands of individual cases. © Springer-Verlag Berlin Heidelberg 2003.