异构平台实时任务的可用性提升容错调度算法

发布时间：2018-04-08 22:36

本文选题：异构平台　切入点：任务调度　出处：《计算机研究与发展》2015年12期

【摘要】：随着互联网+、云计算以及大数据等领域的迅速发展,异构平台成为部署科学计算、工业控制、云存储等关键应用的重要平台.由于平台内处理机性能及软硬件体系结构的异构性,异构平台表现出良好的可扩展性与高性价比.但是平台规模扩大和系统应用日趋复杂导致异构平台上实时任务的可调度性变差,系统可用性降低.针对此问题,提出了一种异构平台实时任务的可用性提升容错调度算法(availability improving fault-tolerant scheduling algorithm,AIFSAL).以处理器利用率和可用性成本为依据设计任务调度整体框架结构、处理机、任务以及调度模型;结合可用性成本设计算法并通过主副版本备份(primary/backup copy,PB)方法实现容错,任务副版本根据处理器利用率不同选择被动或重叠方式执行以减少系统冗余开销,提高可调度性,调度中无论任务主、副版本均优先选择可用性成本低的处理机以提高系统可用性;对任务分配情况和可调度性进行理论分析以证明AIFSAL的可行性.仿真实验与比较分析表明,AIFSAL较可用性约束(availability approached task scheduling algorithm,AATSAL)算法、单调速率扩展(task partition based fault-tolerant rate-monotonic,TPFTRM)算法以及最早完成时间(MinMin)算法在不降低可调度性的基础上有效地提升了系统可用性,减少了系统综合开销,综合性能提高显著.
[Abstract]:With the rapid development of Internet, cloud computing and big data, heterogeneous platforms have become an important platform for deployment of scientific computing, industrial control, cloud storage and other key applications.Due to the heterogeneity of processor performance and hardware and software architecture, heterogeneous platforms exhibit good expansibility and high cost performance.However, the expansion of platform scale and the increasing complexity of system application make the schedulability of real-time tasks on heterogeneous platforms become poor, and the availability of systems is reduced.In order to solve this problem, a availability improving fault-tolerant scheduling algorithm for real-time tasks in heterogeneous platforms is proposed.Based on processor utilization and availability cost, the overall task scheduling framework, processor, task and scheduling model are designed, and fault tolerance is realized by combining availability cost design algorithm and primary / secondary version backup copy PB.In order to reduce the redundant overhead of the system and improve the schedulability, the sub-version of the task chooses the passive or overlapping mode to execute according to the processor utilization, regardless of the task owner,In order to improve the system availability, the secondary version gives priority to the processors with low availability cost, and makes a theoretical analysis of the task allocation and schedulability to prove the feasibility of AIFSAL.Simulation experiments and comparative analysis show that AIFSAL is more efficient than availability approached task scheduling algorithm (AATSAL) algorithm, monotone rate extension partition based fault-tolerant rate-monotonic fault-tolerant TPFTRM algorithm and earliest completion time MinMinalgorithm can effectively improve system availability without reducing schedulability.The comprehensive cost of the system is reduced and the comprehensive performance is improved significantly.
【作者单位】：西安交通大学电子与信息工程学院;
【基金】：国家“八六三”高技术研究发展计划基金项目(2008AA01A202) 国家自然科学基金项目(61202041)
【分类号】：TP302.8

【参考文献】