| 景维鹏,吴智博,刘宏伟,董剑.可靠性代价和Makespan驱动的分布式容错调度算法[J].高技术通讯(中文),2012,22(5):477~482 |
| 可靠性代价和Makespan驱动的分布式容错调度算法 |
| A reliability cost and Makespan driven fault tolerant scheduling algorithm for distributed systems |
| 修订日期：2010-11-19 |
| DOI： |
| 中文关键词: 主副版本, 优先级约束, 高失效率, 主副重叠方式 |
| 英文关键词: primary backup, precedence constrain, higher failure rate, overlapping backup copy |
| 基金项目:863计划（2006AA01A103，2008AA01A201， 2009AA01A404）资助项目 |
| 作者 | 单位 | | 景维鹏 | 哈尔滨工业大学计算机科学与技术学院；东北林业大学信息与计算机工程学院 | | 吴智博 | 哈尔滨工业大学计算机科学与技术学院 | | 刘宏伟 | 哈尔滨工业大学计算机科学与技术学院 | | 董剑 | 哈尔滨工业大学计算机科学与技术学院 |
|
| 摘要点击次数: 3205 |
| 全文下载次数: 2319 |
| 中文摘要: |
| 为解决异构分布式环境下采用主副版本策略的可靠性调度问题，提出一种基于优先级约束的可靠性代价和Makespan（调度时长）驱动的分布式容错调度算法DRCAMD。该算法可在满足系统可调度性的前提下，以异构分布式环境的节点、通信链路的可靠性与Makespan做为可调节局部目标函数，实现具有较高可靠性及较短执行时间的容错调度策略，避免将任务分配到失效率较高的节点上执行。另外，算法的副版本采用被动和主副重叠方式执行，使得容错调度算法具有较大的灵活性。仿真实验表明，该算法性能优于现有容错算法。 |
| 英文摘要: |
| To solve the reliability scheduling problem of primary backup in heterogeneous distributed computing systems, the paper puts forward the DRCAMD, a fault tolerant scheduling algorithm for distributed systems based on priority constraints of reliability cost and Makespan (the schedule length) driven. Under the premise of meeting schedulability, the algorithm realizes a higher reliability and shorter execution time fault tolerant scheduling strategy with the heterogeneous distributed environment nodes, at the same time it can avoid allocating the tasks to the nodes of higher failure rate for execution. In addition, the algorithm of minor version can execute in passive and overlap between main and side, making the fault tolerant scheduling implemented with the greater flexibility. And the simulation result shows that the DRCAMD outperforms the exiting fault tolerant scheduling algorithms. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
