Grid-Oriented Resource Monitoring System

A Grid is a complex system consists of massive distributed heterogeneous resources,thus monitoring and control systems for Grid resources are very important for the resource management and system robustness of Grid systems. This dissertation analyzes the characters and problems of Grid monitoring and control systems, concluding the basic concepts and principles for them, and introduces our two Grid monitoring projects, which cover both theoretic and practical aspects.The characters and problems of Grid monitoring systems are related to the natures of the Grid resources. A main contribution about these is using the internal logic and topology relations of Grid resources to direct the implementations of Grid monitoring and control systems. In practice, the logic relations can help to simplify the resource definitions, while the topology relations can help to partition and combine these monitoring systems.This dissertation analyzes the theoretic problems from several aspects, including function definitions, basic elements, topology structures, and evaluation standards. Especially, two factors, subjects and functions about demands, are used to construct concrete evaluation standards, which are helpful to the design, analysis and evaluation of the Grid monitoring and control systems.Two monitoring and control systems developed by us, GridMon and DTGMS, are introduced. The GridMon is designed for a concrete environment and has been implemented, which uses 3-layer tree structure and unified resource definitions. The DTGMS is a general monitoring system for Grid resources, so it divides all data to three layers, namely abstract type layer, instance definition layer and status data layer, and uses the directed acyclic graph to describe the relations among monitored object instances. Based on the layers of data, the DTGMS consists of maintaining layer and working layer, which fulfill the maintenance of metadata and system, and processing of status data, respectively. The design of DTGMS supports diverse resource types and network topologies, and has strong scalability, extensibility and manageability, which are suitable for Grid monitoring.