Optimal GEDF-based schedulers that allow intra-task parallelism on heterogeneous multiprocessors

Abstract

A variant of the conventional sporadic task model is considered wherein successive invocations of the same task do not have to execute in precedence order. This model is motivated by stream-processing applications where successive data items can be processed independently. The considered hardware platform is assumed to be a heterogeneous multiprocessor with processors of different speeds. Such platforms can be utilized in embedded applications to enable performance guarantees to be made with acceptable energy costs. The main contribution of this paper is to show that preemptive and non-preemptive variants of the global earliest-deadline-first scheduler are optimal with respect to ensuring bounded response times under the considered task model and hardware platform. An experimental evaluation of both variants is presented as well.