Building P2P Volunteer Computing System Using Agents

A P2P volunteer computing system named PPVC is proposed. Its advantages are the flexible topology and the scalability. Volunteers are organized as a P2P network and a decentralized job scheduling method is utilized. It can accomplish an application efficiently under the dynamic changes of the P2P network condition. PPVC is built using an AOSE method and developed on the MAGE environment. The whole system is designed as a multi-agent system and thereby the key functions of the system are accomplished by agent´s abilities and collaborations. Mobile agent is used for job remote execution and intelligent agent for job scheduling. Job splitting is mapped to agent cloning and result composing is corresponding to agent merging. All information exchange follows agent interaction protocols. The work approves the efficiency of the P2P architecture for volunteer computing and the feasibility of the AOSE method.