Q-learning based collaborative load balancing using distributed search for unstructured P2P networks

Peer-to-peer structures are becoming more and more popular and an exhilarating new class of ground-breaking, Internet-based data management systems. Query load balancing is an important problem for the efficient operation of unstructured P2P networks. The key issue is to identify overloaded peers and reassign their loads to others. This paper proposes a novel mobile agent based two-way load balancing technique for dynamic unstructured P2P networks. In this scheme, target peers are selected based on the result of reinforcement learning. Simulation results indicate that our technique manages the load on peers effectively and increases the search performance significantly.