随着云计算技术的日益发展， Linux集群以造价低廉、易于扩充等优势得到了愈来愈广泛的应用。为了更好地发挥集群性能，充分利用集群节点的资源，对集群性能进行实时监控是很有必要的。提出了一种 Linux 集群监控器设计与实现方法。该方法通过每隔一段时间采集节点机/proc虚拟文件系统中的信息，如CPU和内存使用情况等。经过过滤后，通过 socket 传输给监控服务器。论文首先给出了监控器的总体设计方案，整个监控系统由守护在管理节点上的信息管理服务器进程和运行在各个计算节点上的采集器进程组成。然后分采集器和信息管理器两大部分，分别介绍了其具体的设计框架和其采用的关键技术。采集器分主要由信息采集、信息处理和信息传送3个模块组成，分别采用3个线程来完成。信息管理器采用了线程池技术，用以接受采集器发送过来的传输请求。实践证明，该系统可以很好地满足实时监控Linux集群性能的需要。%With the rapid development of cloud computing, Linux cluster is used widely with the advantages of low cost and good scalability. This paper presents a method of design and implementation of Linux cluster monitor. Some information, such as the CPU and memory usage, are collected from/proc virtual file system on every node at regular intervals. After being filtered, these information are sent to monitor server via socket. The overall design scheme of monitor is first described in the paper. It is composed by information management server process, which is run on the management node and collecting processes, which are run on every node. Then their design framework and key techniques are introduced.The collector is composed by three modules that is information collection, information processing and information transmission. These modules are realized by three threads. The thread pool technology is used by information management server, which receives the transfer request sent by collector. Proved by practice, this system can well satisfy the real-time monitoring of Linux cluster performance.
展开▼
