October 22nd, 2015: Resource synchronization protocols and design optimization techniques for Mixed-criticality CPS

  • Speaker: Zhao Qingling (Zhejiang University)
  • Title: Resource synchronization protocols and design optimization techniques for Mixed-criticality CPS
  • Abstract: In Today’s complex Cyber-Physical Systems(CPS), designers often need to integrate multiple applications with varying levels of criticality on the same hardware platform to form a Mixed-Criticality System. Mixed-Criticality Scheduling (MCS) is a type of real-time scheduling algorithm for Mixed-Criticality Systems. In this work, we address two aspects of MCS, including resource synchronization protocols and design optimization algorithms. Firstly, a common assumption adopted by research on MCS is that tasks with different criticality levels are independent and do not share common resources (e.g., global data variables). This assumption is helpful for reducing mutual interference among tasks with different criticality levels and simplifies system design and safety certification. However, for certain cases, tasks with different criticality levels may need to share resources and communicate with each other, especially among the lower-critical levels. Therefore, in order to support resource sharing among tasks with different criticality levels while guaranteeing real-time schedulability, we develop real-time resource synchronization protocols for both Fixed-Priority (FP) and Earliest Deadline First (EDF) algorithms on uniprocessor, and give their properties and schedulability analysis algorithms. Secondly, since automobiles are mass-produced consumer products, the hyper-competitive market pressure forces the automotive industry to be very cost-sensitive, making it very important to reduce system hardware cost by adopting resource-constrained (including CPU speed and memory) low-cost processors. In order to optimize (minimize) system memory resource requirements, we integrate Preemption Threshold Scheduling (PTS) with MCS, for both Fixed-Priority (FP) and Earliest Deadline First (EDF) algorithms on uniprocessor, and develop their schedulability analysis algorithms. In addition, we reduce system stack space requirement while guaranteeing schedulability by optimizing task priorities and preemption threshold values.