集群作业管理系统中负载监视和检查点技术的研究与实现研究-计算机系统结构专业论文.docxVIP

摘要摘要 摘要 摘要 集群系统是一组独立的计算机的组合，他们可以自主的共同协作以完成一 件任务。集群已被广泛应用于高性能计算领域，提供了低成本，可扩展及高性 能的计算能力，在众多的科学计算、工程计算中取得了良好的使用效果。集群 也常被用来提供高可用性的服务，为企业、银行和电信等系统提供高度稳定和 可靠的运行环境。 集群／作业管理系统是构成集群的重要软件系统，它的主要任务是对集群 的资源进行集中的监控和管理，为用户提交的任务分配可用的计算资源，并脏 控和管理作业的执行及结果的返回；同时，他还提供了系统容错和错误恢复的 能力，对于大型的计算任务来讲，可以在事故或错误发生时将其损失减少到最 小程度。 本文主要对集群的负载监视和检查点技术进行了深入的调查和研究，并分 别独立的在实验室环境下实现了集群的负载监视和用于容错的检查点模块，为 研究集群管理系统的实现技术打下了基础。 关键词：集群，负载监视，检查点 llI ^bstractAbstract ^bstract Abstract Cluster system is consisted of a group groups of independent computer nodes， which carl be PC，workstations，even SMP．Every node has the ahility to achieve task by himself；it has its own processor，I／0 device，operation system and everything a single computer has．All these devices carl harmonize themselves to split a single computing task into pieces and dispatch them to every available node，waiting for the end oftask execution and then return tbe result． Cluster system is already widely used in of high performance compute，high available service etc． Cluster and task management system is an important software system which consists of a cluster．Its head function is the centralized managemem of cluster system resource，and also it is responsible for allocating available node for custom’s task， watching every part’S execution and sending result back．To some large scale computing such as weather forecast molecular biology emulation,hydro mechanical computing ete，time to accomplish one task is often calculated by month， cluster managemem system should also provide ability to deal with unpredictable error or accident during the execution oftask． In this dissertation，some technique which cluster and task management system used such as system load and process checkpoint is explanation，also its new design and implement is given．Before finishing this dissertation，some simple optimization and experimentation is shown，indicating validity of work． Keyword：Cluster，Load Monitor，Checkpoint Ⅳ 南开大学学位论文版权使用授权书本人完全了解南开大学关于收集、保存、使用学位论文的规定，同意如下 南开大学学位论文版权