Supporting mode changes while providing hardware isolation in mixed-criticality multicore systems

Abstract

When hosting real-time applications on multicore platforms, interference from shared hardware resources can significantly increase task execution times. Most proposed approaches for lessening interference rely on mechanisms for providing hardware isolation to tasks. However, one limitation of most prior work on such mechanisms is that only static task systems have been considered that never change at runtime. In reality, safety-critical applications often transition among different functional modes, each defined by a distinct set of running tasks. In a given mode, only tasks from that mode execute, yet tasks from all modes consume memory space, and this creates additional constraints affecting hardware-isolation techniques. This paper shows how to address such constraints in the context of an existing real-time resource-allocation framework called MC2 (<u>m</u>ixed-<u>c</u>riticality on <u>m</u>ulti<u>c</u>ore). In MC2, hardware-isolation techniques are employed in conjunction with criticality-aware task-provisioning assumptions that enable hardware resources to be utilized more efficiently.