Adaptive feedback scheduling of incremental and design-to-time tasks

Abstract

This paper discusses an approach for adaptive feedback scheduling in resource insufficient environments. In particular, we examine the problem of maximizing the utilization of the CPU for a collection of periodic incremental and design-to-time tasks with variations in actual execution times. CPU allocation beyond a minimum is performed according to a Quality-of-Service (QoS) based utility function. Schedulability analysis results are utilized to determine guaranteed execution time limits (worst-case schedulability boundary). Past history based on actual task execution times is used to identify the actual schedulability boundary and execution time allocations are adjusted accordingly. This feedback control approach to scheduling supports opportunistic resource allocation beyond the analytic limits, while minimizing deadline misses and limiting them to optional execution increments.