Quality of service capabilities for hard real-time applications on multi-core processors

  • Authors:
  • Jan Nowotsch;Michael Paulitsch

  • Affiliations:
  • EADS Innovation Works, Munich Germany;EADS Innovation Works, Munich Germany

  • Venue:
  • Proceedings of the 21st International conference on Real-Time Networks and Systems
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

Computing Worst-Case Execution Times (WCETs) for applications executed on multi-core processors is a challenging topic since possible interferences on shared resources need to be considered. Some approaches are already proposed in literature, but the problem is still not sufficiently solved. Different approaches suffer different shortcomings. For instance, the mutual analysis of multiple applications leads to great computational complexity, pessimistic assumptions on the interference between tasks causes highly overestimated WCETs and resource privatisation dissipates processor resources. In this paper we tackle the problems of overestimated WCETs due to pessimistic analysis and differences between average-case and worst-case execution timing. We introduce a new computing paradigm for safety-critical real-time systems, which enables Quality of Service (QoS) properties to increase the utilisation of multi-core processors while still guaranteeing bounds on the worst-case behavior. This paradigm is one approach to raise multi-core performance over single-core processors, even for hard real-time systems. For evaluation we use abstractions of real applications. The concepts are implemented on Freescale's P4080 multi-core processor and AbsInt's timing analysis framework aiT. The results show an increased processor core and system utilisation of up to 99.9% and 59.3% respectively, while still providing hard deadline guarantees for all applications.