Mixed-Criticality Systems as a Service for Non-critical Tasks
Mahmood Hikmet, Matthew M. Y. Kuo, Partha S. Roop, Prakash Ranjitkar · 2016
Mixed-Criticality Systems are capable of accommodating tasks of varying criticality. In this paper, these are [life, mission, and non-critical]. Tasks usually have an overestimated execution time to allow for the Worst Case Execution Time (WCET). When these tasks finish execution prior to their allotted execution time due to pessimistic assumptions present in the static analysis of the system. The surplus time is used to accommodate tasks with tolerance for deadline-misses. Non-critical tasks are often treated in a ”best-effort” capacity where no quality of service is considered. When processor utilisation is not overconstrained, all deadlines will be met. However, in cases where not enough processing resources exist to meet all deadlines for non-critical tasks, the allotted time for the non-critical tasks must be rationed between non-critical tasks. This paper proposes a novel method of prioritising non-critical tasks. By treating task execution as a service, non-critical tasks with unbounded deadline miss tolerance are given a Grade of Service for their met deadlines. This Grade of Service is used for the dynamic scheduling of non-critical tasks. Four different scheduling algorithms were tested with the proposed Highest Penalty First algorithm for distributing the effort of task execution amongst non-critical tasks in a proportionate manner and showing superior fairness of task execution compared to all other tested algorithms.