Supporting Predictable Servant-Based Execution Model on Multicore Platforms

Abstract

Since the complexity of real-time systems increases continuously, multicore platforms are widely used in these systems. However, the contradiction between the need for efficiency and predictability in real-time systems has become more and more sharpening. In this paper, we extend a LET-based predictable execution model, named Predictable Servant-based Execution Model, to multicore platforms to make systems both predictable and efficiency. For one thing, PFair based multicore task scheduling algorithm is proposed to properly utilize multicore resources under constraints caused by timing semantics in PSEM. For another, the PFair strategy is extended to be able to schedule tasks with data ependency to make full use of advantages of PSEM. Simulations results on the extended execution model demonstrate the correctness and effectiveness of our design scheme using PFair to map the MPSEM to multicore platform. Besides, our proposal can effectively achieve load balancing and improve resource utilization in multicore real-time systems, without the erosion on the foundation of system predictability.