ABSTRACT

There are a variety of peer-to-peer (P2P) systems for sharing documents currently available. According to their data organization, P2P systems are classified into two categories: structured and unstructured P2P systems. In structured P2P systems, peers are organized according to some mapping techniques, e.g. hashing function. Whereas in unstructured P2P system, peers are connected to each others randomly; resources are not moved to other peers but hosted on site. Unstructured P2P systems offer a more flexible and autonomous environment, as they require less control for the placement of resources and peers. This work focuses on experimenting on unstructured P2P systems. The challenge in unstructured P2P system is designing routing strategies that lead the user in finding the documents needed. Routing strategies in unstructured P2P system need to consider the dynamic aspects of P2P systems; peers are dynamic and constantly joining and leaving the system, network load changes continuously and resources are added and removed over the time. Therefore, the routing strategy must adapt to such changes to maintain its performance. We propose routing strategies that adapt to these changes through learning mechanisms. The learning mechanisms are conducted by observing the internal and external behaviors of the system. Internal behaviors reflect the internal state of peers such as peers’ interest and collection. External behaviors reflect the external state of the system such as network load. In order to measure the performance of the proposed routing algorithms, some common performance measurements are used. These are “response time” and “number of messages generated” or what is commonly referred to as efficiency, “number of answered and satisfied queries” and the “similarity of documents” or what is commonly referred to as effectiveness of retrieval system. The experiment results show that the proposed algorithms are capable of adapting to new changes. By learning to adapt, the system maintains its performance in terms of efficiency and effectiveness. Moreover, comparison with other similar algorithms also shows the superiority of the proposed routing algorithms. Thus, the proposed routing algorithms are good candidates for effective and efficient retrieval of documents in P2P systems.