A two-level random walk search protocol for peer-to-peer networks

Peer-to-peer systems are finding more popularity as a means of sharing large volumes of data between different users. Each node contains a certain amount of data which is typically Mes but can also be of other types such as records In a relational database. There have been different search algorithm that were developed which enable a certain node to find data that Is distributed among the other nodes in the network [2][3]. Each of these algorithms have certain advantages and disadvantages. This paper introduces a search protocol to be used in peer-to-peer networks. It uses a two-level random walk strategy. The querying node selects k(1) random walks with TTL1(time-to-live)=l(1). When the TTL1 timer expires at a particular node, each thread will then generate k(2) threads which will perform k(2) random walks from that node with TTL2=l(2). This protocol has the advantage of increasing the coverage area with the same number of messages and reducing the redundancies which take place when peer nodes repeatedly receive the same query message. On the other hand, the price paid by this strategy is an increase in the average length of the discovered path.