Research On Resource Management System And Resource Information Management

Resource Management, which is considered to be one of the most importantapplications in next generation Internet, attracts much attention among researchersaround the world. The goal of Resource Management System is to integratelarge-scale, decentralized and heterogeneous resources, such as super computers,storage systems, virtual reality systems, and instruments, through high speed networksinto a virtual super computer to deal with large-scale application problems inscientific research, military areas, business application and information service.The core of Resource management System is scheduling various resources thatbelong to each node in distributed environment as a whole. Users can requestheterogeneous and decentralized resources directly and effectively, and all theresources in the system can be used more effectively. The target of Resourcemanagement is to harmonize resources. Resource management system use variousstrategies to realize resources distribution, like optimizing distribution, hierarchicalscheduling, real time adjust, etc. The resource management system should satisfymore kinds of applications, and the distributing strategies can adjust based oncharacters of the applications.The goal of the paper is to integrate decentralized and heterogeneous resources,such as CPU resources, storage resources and network resources. First we design theframework of the resource management system. Then research resource informationmanagement and the other technology in the point.Based on the researching on resource management system and resourceinformation management, we designed and realized a simulation resourcemanagement system. And in the simulation system we researched two techniques.First we research for using resources by the users. A model of prediction to quantifyuser's demand has been proposed. And the quantify of user's next demand waspredicted by the model. The simulation demonstrates that the model could predict thequantify of user's demand well. Then a new task scheduling algorithm (SDM,Similarity Degree Matching) based on prediction to quantify user's demand andsimilarity degree matching was put forward. The demands of both users and theresources were considered in SDM algorithm and the mapping between tasks andnodes could be achieved. The simulation demonstrates that it improve 20% utilization rate of resources and throughput of the whole system.