Multi-Partitioned Single Cores and Predictable Execution of Safety-Critical Tasks | Consortium for Embedded Systems | SIU

Abstract: The benefits offered by the consolidation of functionality in real'time embedded systems are currently counteracted by the need to preserve the criticality requirements of different tasks as they are sharing the same powerful processor (“core”) or platform. These tasks may also be running in  different operating systems (“partitions”) under coordination from a hypervisor. The goal of this project is to investigate and characterize the level of interference between Highly  Safety'Sensitive (HSS) and Less Safety'Sensitive (LSS) tasks running under the same or different operating systems on a single computational core, and accordingly develop scheduling algorithms to maintain the HSS requirements while providing acceptable Quality of Service to the LSS tasks.

Problem: The interference that HSS tasks experience in the presence of LSS tasks and other HSS tasks is challenging. The goal of this project is to investigate and characterize the level of interference between HSS and LSS tasks running under the same or different operating systems (“partitions”) on a single computational core, and accordingly develop scheduling algorithms to maintain the HSS requirements while providing acceptable Quality of Service to the LSS tasks.

Rationale / Approach: The PIs propose to employ a two'stage approach to ensure successful completion of this project. In the first stage, the PIs propose to explore the interference between determinism and responsiveness of individual HSS and LSS tasks executing within the same operating system with specific emphasis on appropriate scheduling algorithms for ensuring the determinism of HSS tasks and the Quality'of'Service (QoS) for the LSS tasks. In the second stage, the PIs propose to explore the interference between tasks running in a real'time operating system (RTOS) and a general'purpose operating system (GPOS) which are both running on a single core with hypervisor support and study the determinism and responsiveness of the operating systems in each case. In this stage of the project, the PIs propose to abstract details of task scheduling within each operating system and focus solely on scheduling across operating systems through the virtualization layer.

Novelty: While multi'partitioned processors have been deployed in practice, they currently do not support dynamic scheduling and resource management. This project seeks to study the effects of such dynamic strategies.

Potential Member Company Benefits: Current practice in industry for safety'critical applications executing in a multi' partitioned processors rely entirely on a'priori scheduling of multiple partitions and tasks therein. The results of the proposed project will serve as a first step towards enabling dynamic task scheduling and resource management in multi'partitioned  processors.

Deliverables for the proposed year: The deliverables for this project are as follows: 1) A comprehensive report detailing the qualitative and quantitative evaluation of different scheduling algorithms in terms of the
achievable determinism and performance of the HSS and LSS tasks. 2) A comprehensive report detailing the interference of GPOS and RTOS on multi'partitioned cores with hypervisor support. Modified hypervisor and operating system source code, if any.

Milestones for the proposed year: Q1: Exploration of existing research in the area of mixed'criticality systems and development/evaluation of new scheduling algorithms. Q2 & Q3: Characterization of HSS and LSS tasks executing in multiple partitions on a single core with hypervisor support (XEN, KVM). Q4: Report writing and technology transfer.