Research On Fault-tolerance Method For Mobile Agent

In recent years, with the popularization of Internet and the flourishing development of mobile computing technology, the research upsurge on mobile agent is launched not only in academe but also in industry. Different from the existing paradigms like message passing, remote procedure calls, and distributed objects, mobile agent paradigm offers more properties. Mobile agent can act autonomously on behalf of a user and travel through a network of heterogeneous machines, which also leads to many problems for the agent's execution. The mobile agent can be prevented from proceeding with its execution by some reason. At present, some valid approaches to this matter have been proposed. For instance, replication accepted popularly can solve the blocking problem successfully. But it is followed by another problem-because it is very difficult for us to find a correct failure detection in asynchronous system such as the Internet, the mobile agent fails and recovers later, they all lead to a violation of the exactly-once property for the agent's execution.In this thesis, author firstly outlines the current research status at home and abroad, introduces some relative theories. Then, analyzing the advantages and disadvantages of current fault-tolerance model for mobile agent, we present a MAFT model based on mobile agent as a solution for these disadvantages. We introduce a component called Fault Tolerance Module (FTM) in our model. It is agent dependent so that we can achieve fault-tolerant mobile agent execution without modifying existing mobile agent platforms. At first, we present fault-tolerant mobile agentexecution as a sequence of agreement problems; Then, we design a DIV consensus algorithm in order to solve the agreement problems; At the same time, we give a pipelined mode for reliably forwarding the agent between stage s\ and si+i. In addition, we design a CM for every stage of agent's execution in order to record all agent replicas' location, status and fault-tolerant information. By the interaction of the CM, the FTM and user-defined agent we can achieve nonblocking and exactly-once execution of agent successfully at the same time. Then author makes some simple simulative experiments about our fault-tolerance mobile agent execution model. At last, the paper puts forward the summarizing of this thesis and next step's work.