Real-time task pipelines allocation on multicore systems

An important class of real-time multimedia streaming applications can be modeled as pipelines of tasks to be executed on multi processors systems. Each pipeline is periodically activated and each instance has to be completed before a given end-to-end deadline. A general problem is to allocate multiple real-time task pipelines on multi-core systems, guaranteeing their schedu- lability and optimizing a user defined objective (e.g. minimum number of processors, minimum energy consumption, etc.). The objective of this paper is to analyze the problem from a mathematical point of view, and to design an algorithm to explore the space of possible solutions. The research activity builds upon a previous investigation about the assignment scheduling parameters of the tasks. In this thesis, we propose a grouping strategy that tries to minimize the number of processors required to schedule a sets of pipelines, and hence the overall resource utilization. The proposed algorithm is evaluated on a wide combination of parameters and its effectiveness is compared against existing solutions.