| Grid is a new Infrastructure which will have much more powerful computing ability and storage ability than any high performance computer in the world. The grid, which holds various heterogeneous resources distributed in the global field, can provide transparent remote access, resource sharing and distributed computing to the user. Because the heterogeneity and distributing of the grid environment, Monitoring system is very important to the grid. Grid monitoring system can be used in fault detection and recovery, system bottle-neck analysis and system tuning; also, the job scheduler can get information from monitoring system and then select low loaded node to keep the load balance. In present, many organizations and companies are working in the grid monitoring field.The Grid Monitoring Architecture (GMA) is a memo which is proposed by the Performance Working Group of Global Grid Forum. The GMA defines the standard terms and designs a producer/consumer model to regulate the implementation of grid monitoring systems.Taken fully consider of the users' demands, we design and implement a Grid monitoring System (shortly GMS) which based on GMA. The GMS is designed in a layered architecture and the whole system is formed with collective layer, aggregate layer, intermediate layer, application layer and service layer. This kind of layered architecture makes the system easy to scale and add new function to system. Also, this design makes the system more robust and can avoid the single point failure. GMS use sensor plug-ins to collect performance information. User can implement custom sensor to collect the interested information or remove the uninterested sensor from the system. GMS overcomes the shortcoming of traditional monitoring system which can only collect the predefined data type. In GMS, we implement a kind of performance forecasting component which can be used in providing a short time forecast for system performance. This can help the scheduler to get load balancing. GMS takes... |
PDF Full Text Request