基于LDAP的网格监控系统

A LDAP BASED MONITORING SYSTEM FOR GRID

对于高性能分布式计算环境——网格——来说 ,监控其中计算资源的状态是至关重要的 .通过监控可以及时发现并排除故障 .通过分析监控数据可以找出性能瓶颈 ,为系统调整提供可靠的依据 .Grid Mon是基于 L DAP目录服务的分布式网格监控系统 ,改变了以往目录服务中不存储动态信息的使用方法 ,灵活地将静态和动态信息结合在目录层次中 ,从而减少了客户端对服务器的交互次数 ,并采用中间件技术有效地解决了直接访问被监控主机带来的安全和接口问题 .借助 L DAP的目录层次 ,建立了网格系统的树状基本结构 .提出了网格监控对象和监控事件的概念及其表示方法 ,从而形成完整的网格监控结构模型 .详细讨论了根据这个模型实现的网格监控原型系统—— Grid Mon.最后 ,通过网格与机群系统的结构不同点 ,阐述了评价网格监控系统的要点 ,并以此为依据 ,结合应用前景对 Grid

Monitoring the computing resource is very important to the High Performance Distributed Computing Environment-grid. Hardware and software failure can be found and solved in time by monitoring the system. Analyzing the data gathered by monitoring will help find performance bottleneck, which is important for improving the system performance. GridMon is an LDAP based distributed grid monitoring system. It changes the normal usage of LDAP that doesn't store dynamic information in directory service. GridMon stores not only static data, but also dynamic data to reduce the interaction between client and server since the data can be obtained in one access. The monitoring object location and monitoring event are combined together into the same directory service. GridMon uses middleware technology to avoid direct access to the monitored hosts, therefore solving security and interface issues. Based on the LDAP's directory hierarchy, a tree-like infrastructure is built to map the grid structure. The whole grid monitoring system is constructed by giving the concepts of grid monitoring objects and monitoring events, with the representation method. The implementation issues about an implemented system-GridMon, is also discussed, which is designed according to the conceptual monitoring object and event model. The main software architecture used in GridMon is discussed by referring to the data model and design. Finally, the structure deference between grid and the cluster system is analyzed to explain the key points of grid monitoring, and GridMon is evaluated by these points and its potential applications.