Scheduling Constrained-Deadline Tasks in Precise Mixed-Criticality Systems on a Varying-Speed Processor
Scheduling Constrained-Deadline Tasks in Precise Mixed-Criticality Systems on a Varying-Speed Processor
复制标题
DOI：
10.1145/3534879.3534897
复制
发表时间：
2022-06
期刊：
影响因子：
--
通讯作者：
Tianning She;Zhishan Guo;Kecheng Yang
中科院分区：
文献类型：
--
作者：
Tianning She;Zhishan Guo;Kecheng Yang
实时系统通常需要在所有可能的情况下，包括最坏的情况下的保证。因此，当每个系统参数由单个估计指定时，不可避免地会引入显着的悲观情绪。为了缓解这种悲观情绪，已经提出了混合临界（MC）设计，其中一个单一的系统参数提供多个估计。MC调度的大多数现有的工作是针对（全部或部分）牺牲低关键的工作负载的情况下，高关键的工作负载超出正常情况下的估计。最近，另一种称为精确MC调度的方法已经被研究，在这种情况下，没有低关键工作负载被牺牲，但是处理器的速度被提升以适应高关键工作负载的额外执行要求。精确MC调度以前的工作集中在隐式的最后期限的任务。在这项工作中，我们扩展了精确的MC调度的努力，以约束的最后期限的任务，通过开发基于需求的可扩展性分析，在以前的工作中，基于利用率的。这种新的分析还可以实现更灵活的虚拟截止日期设置。合成实验表明，通过这种新的分析和设置虚拟截止日期的灵活性，实现了显着的可扩展性改进。
Real-time systems usually require guarantees in all possible scenarios including the worst case. As a result, when each system parameter is specified by a single estimate, significant pessimism is inevitably introduced. To mitigate such pessimism, mixed-criticality (MC) design has been proposed, where a single system parameter is provided multiple estimates. Most existing work on MC scheduling is directed to (fully or partially) sacrifice low-critical workloads in the event of high-critical workloads overrunning their normal-case estimate. Recently, another approach called precise MC scheduling has been investigated, where no low-critical workload is sacrificed in such situation but the speed of the processor is boosted to accommodate the extra execution requirement by high-critical workloads. Prior work on precise MC scheduling has focused on implicit-deadline tasks only. In this work, we extend the efforts in precise MC scheduling to constrained-deadline tasks by developing demand-based schedulability analysis in place of the utilization-based ones in prior work. This new analysis also enables more flexible virtual-deadline settings. The synthetic experiments have shown that significant schedulability improvements are achieved by this new analysis and by the flexibility in setting virtual deadlines.