A popularity-based globally structured hybrid peer-to-peer network

The hybrid peer-to-peer system combines structured and unstructured peer-to-peer networks. Structured networks provide an efficient lookup mechanism by means of DHTs, but the maintenance of DHT is costly. In contrast, unstructured networks are flexible under the dynamic environments, but perform poorly for locating rare items. The hybrid peer-to-peer system utilizes advantages of these two types of networks. In this work, we focus on globally structured hybrid system. A structured network forms the backbone of the system and multiple unstructured networks are attached to each backbone node. We show that if items are inserted into the system according to their popularity, search efficiency will be considerably improved. We discuss the system architecture of the proposed hybrid network and insertion and search algorithms in this system. The proposed architecture is studied through simulation. Experimental results demonstrate the efficiency of the proposed method.