Research On Resource Monitoring Problems In Grid Environment

Grid [1] is a new type of network computing platform, which is built on the Internet as a new type of information technology infrastructure.Grid integrates high-speed Internet, high-performance computers, large databases, sensors, remote equipment into a whole, in order to share the resources such as computing resources, storage resources, communication resources, software resources, information resources, and knowledge resources, etc. Thus, It is an inevitable choice to achieve the full share of the resources in grid environment.However, due to grid resources with a distributed, dynamic, heterogeneous and autonomous characteristics, as well as the computing resources may be distributed throughout the world, and be connected through the Internet, which makes the effective management of grid resources more complex than the traditional distributed network. Because of these characteristics, it will be some problems such as node failure in the running, the disconnection of the network, or performance problems. Meanwhile, the action that the nodes join and leave the grid environment dynamically, will also make the grid resource management be more complex. This requires to provide a resource monitoring mechanism to be responsible to monitor all kinds of static and dynamic resources, in order to collect the status change information of a variety of resources and the nodes, allow that users and applications will be informed of the information timely,such as resource allocation and scheduling, network bandwidth, processor load, the system throughput, and other information, and timely resolve the obstacles of the grid system in order to improve performance of the entire grid. Thus, in the high-performance computing environment, such as grid, monitoring is very important in the grid.This paper analyzes the domestic and international research, difficulties and needs of the grid monitoring, and summarize the deficiencies of the current Grid Monitoring Architecture Model. On this basis, according to the specific circumstances of the grid, this paper regard the extended structure of tree as the Grid Monitoring Architecture Model, and presents a good service-oriented dynamic Distributed Hash Table resource discovery strategy. On the basis of the previous two, this paper proposes a grid resource monitoring mechanism– Grid Resource Monitoring, research the logic implementation and technology part of the mechanism of the Grid Resource Mechanism, and have done the research and experiments to monitor the relevant resource information. Finally, through experimental tests, performance analysis shows that the mechanism is safe, reliable, effective and can meet the requirements of the grid resource monitoring application. Specifically, this article has done the main work and innovation points including:1. According to the characteristics of grid, a reasonable grid resource monitoring architecture is established. By investigating a lot of literatures, this paper compared Grid Monitoring mechanism widely used at present. It summarizes the deficiencies of the Grid Monitoring Architecture Model used at present, and analyses the improved methods simply. This paper uses the extended structure of tree as as the Grid Monitoring Architecture Model.In this model, it divides grid nodes into domains, and unifies the service of nodes. This structure has good scalability, and it is loose coupling between the various parts in monitoring system, in order to deploy easily in a distributed environment.2. Research on Grid Resource Discovery Mechanism. In this paper, it defines the registration process of node resources in the grid, and designs a resource discovery strategy dynamic Distributed Hash Table based on the extended structure of tree in grid environment. The strategy focuses on how to conduct the discovery and monitoring process of dynamic resource, with the use of it, you can not only accurately monitor the real-time situation of grid resources on the nodes, but also it is dynamic strongly, so as to be able to adapt to the dynamic changes of grid resources.3. Plug-in sensor management[2]. Traditional monitoring system generally only collect a few pre-defined monitoring data type, so it does not meet the needs of different users. In this paper, it uses the technologies such as Web Services, Java Native Interface etc. to design practical plug-in sensor components, so that customers can define your own plug-ins according to their needs to gather useful information from the system or remove a plug-in to lift useless information, which will overcomes the shortcomings that the traditional monitoring is not easy to collect user-defined content.4. Logic and technical implementation of the Grid Resource Monitoring mechanism. This part is experiment, it uses the dynamic Distributed Hash Table algorithm on the basis of the monitoring architecture in the first part and the extended tree in the second part. After finding the resources in grid environment, this chapter focuses on the analysis and implementation about the two aspects of logic and techniques about the Grid Resource Monitoring, and describes remote registration, discovery and monitoring management process of detection of dynamic join / leave resource node. At the same time, this section describes the plug-in sensor technology, with the introduction of concepts of data collection probes, the dynamic adjustment method of monitoring interval, and monitors the detected resources in order to display them to users in the form of graphic vividly. Finally, it is proved the Grid Resource Monitoring mechanism is effective and available by experiment, and analyzes the experimental results.