Research On Implementation Mechanism And Resource Management Of Metacomputing

With the constant enlargement of the scientific calculation scale,large scale computation is also applied more and more in all kinds of business. People's demand for computing capacity is infinite, although the rate of CPU increased rapidly according to Moore's Law — double every 18 months, some large scale computation problems cannot be completed by a PC in a reasonable amount of time. Super Computers are very expensive, and they are usually owned exclusively by some university or large corporation, ordinary users usually have no chance to use the Supercomputers. It is not economical for ordinary organization or person to purchase advanced computational equipment to meet its demand for computation. The number of computers that are connected to Internet including super computers increased greatly , and most of the computers are much underutilized during most time, the concept of metacomputing is introduced.Metacomputing is also known as grid computing, it is a network that connects all kinds of geographically distributed computers(including clusters), database and other equipments, which provides a transparent virtual high-performance computing environment to uses. By metacomputing, resource sharing in a large area and cooperative work are realized, the application of metacomputing includes scientific and engineering computation, information service and cooperative engineering,etc.This dissertation's research focuses on the implementation mechanisms in the application field of information service and scientific computing. In the information service field, from a service-oriented viewpoint , a general integration model of Web service is presented which make use of existing basic Web services, and the model provide a alternative implementation mechanism for metacomputing. In the field of scientific computing, sharing of computing resource is realized by using other's computing equipments to run one own application. To take the challenge of resource management and task scheduling in metacomputing,using existing related research achievements for reference, an improved model of resource management based on market is presented which can not only regulate the supply and demand of computing resource, but also stimulate more computing nodes to enroll in the metacomputing to contribute their computing power, leading the benign devolepment of metacomputing rotatively. A kind of ant algorithm for task scheduling is designed to get better response time and higher utilization of system in metacomputing. Finally, metacomputing sytem based on the presented resource management model is designed, some difficult points of its prototype is given detailed reasearch.