Research Of Grid Monitoring Based On Ganglia And MDS

Grid resources are distributed, heterogeneous, and the number of large and random changes, the grid nodes may join and leave the grid environment at any time, these features will lead information resources and tasks to run frequent collection, and consume a lot of node resources, but also easily lead to the resources assigned to the invalid nodes in grid task scheduling, and seriously affecting the implementation efficiency of grid task scheduling. So, to build a flexible and efficient grid resource monitoring system, monitoring the node resources and consumption information in grid task of running, is an important aspect of grid computing.Campus computing uses computing resources in the campus grid built for a variety of research within the campus environment to provide high-performance computing power. The core technology of Campus Computing Grid (UCGrid 3.0) is based on GT4 which is a distributed computing grid middleware, and as a basis for the establishment of the campus computing grid portal, it provides the campus computing grid to run a variety of function.Campus computing grid can solve the problems that many resources on campus not be fully utilized and at the same time many tasks cannot get the appropriate resources. The building of the campus computing grid must first solve the resource monitoring and resource discovery problem. Resource monitoring is the basis of resource discovery and resource allocation, which provides the necessary resource-related performance data for the campus computing grid.In this thesis, the monitor object is campus computing grid platform UCGrid3.0, this thesis designed and developed the campus computing grid monitoring system, including the campus computing grid monitoring portal and the campus computing grid monitor. Previously research on the campus computational grid monitoring system only limited to resources monitoring, this study includes monitoring system for monitoring and control in one, and proposed a comprehensive monitoring program for the campus computing grid.Campus computing grid monitoring portal is a real-time monitoring system which monitors a variety of resources of campus computing grid, it provides CPU, memory, hard drives, network traffic and other static and dynamic resource information in real-time monitoring through the graphical show. To make up for GT in the MDS component of the comprehensive monitoring of resource information and real-time gaps, the campus computing grid monitoring portal which based on Ganglia secondary development introduces the Ganglia monitoring framework. Monitoring portal communicates with gmond and the grid resource real-time information which collected by gmond would be shown through the web page, while real-time resource information is written to the MDS and database, it provides strong resource data support for grid scheduling.Campus computing grid monitor CGAMS can monitor the campus computing grid middleware, GT grid environment, grid database etc, it makes the campus computing grid to run persistent. CGAMS is divided into the central control-side and client feedback-side, they are deployed in the center node and sub-nodes of the campus computing grid respectively. Both use the asynchronous communication mechanism based on Socket to communicate. Central control-side monitors campus computing grid middleware, and GT components monitor on central node receives feedback-side to send GT component running status information, meanwhile the central control-side canrestore the sub-node's GT to re-work. CGAMS using multi-threading technology, ensuring it can be in parallel to complete a variety of different functions at the same time.Composited by the campus computing grid portal and grid monitor, the campus computing grid monitoring system not only real-time monitors a variety of resource information in grid environment, but also can control a variety of components related to campus computing grid. it provides a set of monitoring and control in a comprehensive resource monitoring solution。 Campus computing grid monitoring portal and grid monitor CGAMS have deployed on the existing campus computational grid environment. The monitoring system has achieved a good result in the running, not only to provide a comprehensive data support for grid resource scheduling, but also to ensure the campus computing grid to run persistent.