Quantitative Analysis of ARINC-653 Scheduling Overheads on Multi-Core Systems

Abstract

The Integrated Modular Avionics (IMA) architecture has been suggested to address the Size, Weight, and Power (SWaP) issues and to provide better software consolidation and testability by means of partitioning. ARINC-653 defines essential software features for temporal and spatial partitioning and APplication EXecutive (APEX). The multi-core systems can cope with the SWaP issues efficiently, but multi-core processors have side-effects because of the hardware resources shared across multiple cores. In this paper, we aim to quantitatively compare two different implementation alternatives of ARINC-653 scheduler named Centralized ARINC-653 Scheduler (CAS) and Distributed ARINC-653 Scheduler (DAS) on a multi-core system. It is important to minimize the scheduling overheads to stably utilize existing hard real-time scheduling algorithms; however, the discussions about ARINC-653 scheduling on multi-core systems were limited to qualitative analyses in literature. Thus, we quantitatively analyze the scheduling overheads of CAS and DAS and provide insights into critical design issues of the ARINC-653 scheduler on multi-core systems.