Gang fixed priority scheduling of periodic moldable real-time tasks

In the past decade, the development of multicore and multiprocessor platforms attracted a lot of attention in the real-time community. At first, researchers considered several ways to take advantage of this kind of platform to process sequential tasks. Over the past few years, people started considering parallel tasks to take even better profit of this new architecture, or to save energy. There are several families of parallel tasks. One of them is the “Thread” model, where threads can run (rather) independently [4]. Another one is the “Gang” family, requiring that all threads start and stop using the processors synchronously. Tasks can then be seen as rectangles, where the width represent the time, and the height the number of processors. As presented in [2], the Gang task family can be split in three sub-families: rigid tasks [2, 3], Malleable [1] and Moldable. In the latter, considered in this paper, the scheduler chooses the number of processors for each job1, and this number does not vary with time for a given job (but may be different for two jobs of the same task), we propose algorithms and sufficient schedulability conditions in this research. As far as we know, this is the first paper studying moldable tasks for real-time systems.