Optimizing Routing Performance in P2P Networks Using Machine Learning

Peer to peer (P2P) networks use blind routing strategies, as there is no pre-knowledge about presence of the nodes in vicinity. Due to this blind or adhoc routing nature of P2P networks, generating and maintaining routing tables is generally done for every source destination pair which is communicating data with each other. Due to this fact, a number of routing algorithms use beacon-based routing, where the sender sends a routing request, while the neighboring nodes reply with an acknowledgement and other node specific parameters, so that distance and other calculations can be done for an effective routing process. In this paper, we propose a machine learning based P2P routing protocol (MLPR) which uses bio-inspired computations for route evaluation. The proposed protocol is compared with the standard P2P routing techniques like Viceroy, Tapestry and Kademlia in order to evaluate the Quality of Service (QoS) superiority, and it shows that the proposed algorithm is nearly 10% superior in terms of end-to-end delay, energy consumption and packet delivery ratio than these aforementioned techniques.