Blocking-Aware Partitioning for Multiprocessors*

In the multi-core and multiprocessor domain there are two scheduling approaches, global and partitioned scheduling. Under global scheduling each task can execute on any processor while under partitioned scheduling tasks are allocated to processors and migration of tasks among processors is not allowed. Under global scheduling the higher utilization bound can be achieved, but in practice the overheads of migrating tasks is high. On the other hand, besides simplicity and efficiency of partitioned scheduling protocols, existing scheduling and synchronization methods developed for uniprocessor platforms can more easily be extended to partitioned scheduling. This also simplifies migration of existing systems to multi-cores. An important issue related to partitioned scheduling is how to distribute tasks among processors/cores to increase performance offered by the platform. However, existing methods mostly assume independent tasks while in practice a typical realtime system contains tasks that share resources and they may block each other. In this paper we propose a blockingaware partitioning algorithm to distribute tasks onto different processors. The proposed algorithm allocates a task set onto processors in a way that blocking times of tasks are decreased. This reduces the total utilization which has the potential to decrease the total number of needed processors/cores.