摘要
随着服务器系统节点数量和用户任务的不断增多,使得风暴节点出现冲突问题。在此背景下,通过分析服务器系统的节点连接架构,将检测器部署在其通信平台上,对用于检测服务器多冲突风暴节点的MooseFS检测器进行设计与实现。当风暴节点中存在较多冲突时,MooseFS检测器采用动态冗余分布策略与硬件平台协同检测,冲突不多时只使用硬件进行检测。动态冗余分布策略选用冗余度度量节点任务开销,给出风暴节点子任务、带宽和中央处理器运行时间的可用度函数,检验冗余度是否可靠。硬件平台通过逻辑门处理器对子任务、带宽、中央处理器运行时间以及存储容量的负载率进行度量。实验结果表明,所设计的MooseFS检测器拥有很强的资源控制能力。
As the number of server system nodes and user tasks increases unceasingly, conflict of storm nodes occurs. Un- der this background, the MooseFS detector used for detecting multi-collision storm nodes in servers was designed and implement- ed based on the analysis of the node connection architecture of the server system and the deployment of the detector on the com- munication platform of the server system. When many conflicts exist in storm nodes, the dynamic redundancy distribution strate- gy and hardware platform are used for the MooseFS detector to execute collaborative detection. When not many conflicts exist, only hardware is used for detection. In the dynamic redundancy distribution strategy, redundancy measurement task node over- head is selected, and the availability functions of storm node subtasks, bandwidth, and CPU running time are given to detect the redundancy reliability. On the hardware platform, the logic gate processor is used to measure subtasks, bandwidth, CPU run- ning time and load rate of storage capacity. The experimental results show that the designed MooseFS detector has strong re- source control capability.
出处
《现代电子技术》
北大核心
2017年第18期26-28,共3页
Modern Electronics Technique
基金
国家自然科学基金(71203229)
北京市教委科技计划面上科研项目(KM201310857001)
