The Research On Grid Resource Trading And Recommendation Model

Grid is a virtual and high-performance computing environment which is relatively transparent to users. It connects and integrates huge amounts of geographically distributed and heterogeneous resources through Internet with the purpose of achieving resource sharing globally. Owing to its great application prospect, Grid has become a hot-spot in recent years. Resource Management is the kernel of a grid system and has direct impact on system performance. With the widely use of economic models in this field, resource trading is turning into a new research branch of resource management.Aiming at making up the deficiencies in the aspect of payment mechanism of existing resource trading models, a new universal resource trading model was porposed followed by two optimization schemes.1. A Resource Trading Model Based on Third-party Payment Mechanism. By introducing a third-paty to manage and control the payment process, the new model can insure secure tranactions. A crisis management mechanism was also provided in the model in order to identify the cheating actions of both the trading parties. Analysis of the model indicates that the new model can help to reduce the risk during the transaction process and maximize the economics benefits of both providers and consumers.2.A Resource Cluster Method Based on Maximal Theta-distant Subtrees. The main idea of this point is clustering preprocessing which devides all the resources into clusters with different performance based on the performance distance between various resources. By doing this, the grid system only need to search in the cluster that most close to the user's demand instead of whole resource pool when getting resource lists for further resource allocation. Consequently, cluster preprocessing can greatly reduces the time consuming of resource matching and improve the efficiency of resource allocation.3. A Resource Recommendation Algorithm. The algorithm was proposed in order to balance the system load caused by the variation of user demands. With the help of recommendation algorithm, users are albe to make tranactions with the resources recommended by grid system directly without the on-line allocation of grid system when the system is busy. By implementing resource preallocation, recommendation mechanism can effectively reliefs the system pressure during apply peak and reduces users'waiting time.