Extending Static Scheduling Algorithms for Moldable Tasks towards Dynamic Scheduling of Multiple Applications with Soft Real-time Properties

摘要

We investigate scheduling of multiple parallel applications with real-time constraints and the goal of energy minimization onto a multicore processor with frequency scaling. We propose a method to employ recent static scheduling algorithms for parallelizable tasks and adapt them in order to create a dynamic scheduler for the situation where applications may terminate, new applications may start, or applications change their behaviour, e.g. their degree of parallelism, at runtime. We list achievements so far and open problems.