基于虚拟计算环境的核心机理,提出由自主调度单元、域调度共同体、元调度执行体为核心的多机群协同系统框架.剖析多机群任务并发运行性能模型,设计了多机群协同调度算法框架,提出最大空闲节点优先、最小网络拥塞优先、最小异构因子优先与最小异构空闲节点优先4种启发式资源选择策略.实验验证了协同调度模型与算法在任务集完成时间与系统平均利用率的测度上的有效性.
Based on the core mechanisms of Internet-based virtual computing environment (iVCE), a novel architectural framework for the multi-cluster task co-allocation is proposed by introducing the autonomic scheduling elements, domain scheduling commonwealth and meta-scheduling executor. A new multi-cluster task scheduling schema based on the multi-cluster task execution performance model is presented. Four multi-cluster heuristic scheduling algorithms are provided. Experiments indicate the scheduler schema and the algorithms are effective in the objective function of makespan and average utilization.