论文部分内容阅读
为了实现机群系统内负载的均衡分布 ,提高资源利用率和系统的吞吐率 ,论文设计并实现了基于并行虚拟机环境的可迁移动态负载平衡系统。该系统采用集中式任务调度 ,定义了负载向量的计算方法 ,同时设计了“自适应双阈值算法”实时监测系统下各结点的负载信息 ,并实现了同构机群系统下进程的动态迁移。该算法有效保证了负载信息的及时性和准确性 ,避免了进程迁移的颠簸现象。实验结果表明 ,由于采用了自适应算法 ,本系统额外开销较小 ,能有效改善动态任务的分配和调度 ,提高系统的并行运行性能。
In order to achieve a balanced distribution of load in the cluster system and improve resource utilization and system throughput, the paper designs and implements a migratable dynamic load balancing system based on parallel virtual machine environment. The system adopts centralized task scheduling and defines the calculation method of load vector. At the same time, the load information of each node under the real-time monitoring system of “adaptive double threshold algorithm” is designed, and the process of dynamic migration of isomorphic cluster system is realized. The algorithm effectively guarantees the timeliness and accuracy of the load information and avoids the bumpy phenomenon of the process migration. The experimental results show that due to the adaptive algorithm, the overhead of the system is small, which can effectively improve the allocation and scheduling of dynamic tasks and improve the parallel operation performance of the system.