Meeting hard-real-time constraints using a client-server model of interaction

Abstract

The Ster environment integrates programming languages with scheduling strategies to support the development of flexible and predictable distributed systems. Ster application objects can establish a client-server relationship, and clients may specify sequences of services that should be treated as atomic multiservice accesses. The core of Ster's scheduling strategy is an offline scheduler that searches for a schedule that guarantees application's allocation, timing, precedence and resource constraint and tries to balance the schedule in the space and temporal domains. The experimental evaluation of the offline scheduling algorithm discussed, shows that the simple heuristics and mechanisms it implements are very effective in finding well balanced schedules.