Mobile Agent-based Parallel Computing Research

With the need of science and engineering computing increasingly, more and more people hope to use the environment of parallel computing. However, except few large computer users, most people have no chance to use parallel computer for their task due to various limitations of conditions. In addition, with the development of network technology, abundant resource permeated the whole network. It has become a problem drawing attention that how to use the disengaged resource availably. Under this background, cluster system has come into being. Because of its convenient programming, low investing venture, high performance-price ratio, flexible architecture and well scalability, cluster system has being a focus in parallel computing, but many problems still exist such as complex programming, poor portability. At the same time, the features such as low bandwidth and high latency have been becoming the bottleneck limiting the development and popularization of network. To solve it, after many computing model introduced, a new one , mobile agent was proposed. Mobile agent is becoming a focus in network computing characterized by its mobility, autonomy, reactivity, asynchronism, cooperation and being superior in alleviating network load, asynchronous and autonomic computing, parallel processing, etc. In this thesis, after analysis and research of the mobile agent technology and parallel computing technology deeply, author introduced an agent-based parallel computing model and discussed questions such as communication mechanism between agents, fault-tolerance mechanism, load-balance mechanism based on this model deeply. The major content include the concept and principle of mobile agent, the concept and principle of parallel computing, the feasibility of mobile agent integrating with parallel computing, the model of parallel computing based on mobile agent and so on. In the end of this paper,author analyze the advantages of parallel computing agent-based compared with traditional parallel computing.