The Component Moniroting Model In GridCOMP

Grid computing has become a hot research area in recent years due to its objective of resource sharing across administrative domains. In this dissertation, we focus on grid monitoring systems, which collect and deliver runtime status information of grid resources to achieve a more reasonable scheduling. Design of such a system will face new challenges arising from the special characteristics of grid resources, such as, heterogeneity, autonomy, dynamic, etc. Several key problems in the grid monitoring system are studied here, including grid monitoring model, monitoring data discovery and location, and real-time collection of monitoring data. We also propose corresponding approaches for them.Main contributions of this dissertation are as follows:(1) We present a JMX-based monitoring component model in GridCOMP. It is a grid resource monitoring system that supports automatic component level monitoring in grid environment. With its help, we can improve the QoS of grid resource sharing and provide a set of uniform access interfaces to grid monitoring data. Furthermore, this model makes it possible to leverage decision-making load across administrative domains, and perform the monitoring of grid components along with the close-loop structured resource monitoring.(2) The monitoring model is designed as a primitive component, which composes a composite component with the monitored component by using GridCOMP.(3) The composite monitoring component can be deployed on the remote JVM(Java Virtual Machine) using XML (The Extensible Markup Language) technology of ProActive.(4) Our design benefits from dataflow techniques to deal with the real-time monitoring data and the various query requests. In addition, we make a further study of the query optimization for the monitoring data. (5) The model presents a method of calculating resource consumption, which is implemented with the non-recursive traversal algorithm that uses tree to store system process tree.