Real-time scheduling for nested-parallel task model on multi-core processors

In today's scenario, lots of real-time applications are encountered which includes numerous computations that have to be done in parallel in stipulated period of time. For such tasks single core processors cannot perform well even multiprocessors are not able to, because traditionally, sequential models were used which ignore parallelism between the tasks. However, parallel models have the capability to parallelize specific segments of tasks, thereby leading to shorter response times as possible. In this paper, the problem of scheduling the real time, periodic, implicit deadline nested-parallel tasks on multi-core processors is discussed. Initially, a nested-parallel task model, which is synchronous in nature wherein each task consists of number of segments and each segment consists number of parallel or nested threads, is considered. After, a novel task disintegration technique that disintegrates each parallel or nested-parallel task into a set of sequential tasks is proposed. Here it is proved that the proposed task disintegration technique achieve a resource augmentation bound of 3.414 when scheduled disintegrated tasks using global EDF scheduling, which means, if any task set is feasible on m unit speed processors, it can be scheduled using proposed technique on m 3.414 times faster processors.