Management Strategy Based On Domain Index In P2P Network

With the continuous expansion of the Internet both in function and in scale, the burden of the server has become heavier and heavier. The defect of poor efficiency and being difficult to expand in the client/server model exposes gradually, which makes the C/S model hard to adapt to the modern Internet which has a higher and higher requirement to the efficiency. The peer-to-peer network, namely P2P network, is a combination of Computer Networks and Distributed Systems. In a P2P network, clients’reliance on the server is weakened, as there are no differences between clients and servers, in other words, all peers in the network are equal, free and interacted, which share files and deliver messages directly with each other. Based on this characteristic of P2P network, we can distribute the replicas of a file to different peers reasonably, and thus to improve the efficiency of the network. However, the characteristic of free and equal P2P network leads to some drawbacks, such as frequent churn and distribution of information, which affect the management of the replicas and the maintenance of the consistency of the replicas negatively.We make an intensive study of the management of the replicas in P2P network in this paper. Even though, a P2P network works well in most of the time, it may cause many problem of this kind or that kind during the actual operation lead to the dissatisfaction of client peers’ needs, such as the problem of balancing the loads of different peers, namely, the "hot spot" problem caused by the heavy burdens of some peers and the "cold spot" problem that leads to the difficulty in searching. There is also the problem of maintaining the consistency of replica after data modification and data update, or caused by peers’existing and joining repeatedly in the P2P network. Due to all these problems above, we must make a further research on the strategies of replicas management so as to improve the availabilities of the replicas in the P2P network.To solve the defects in existing replicas management strategies, which lack a comprehensive consideration of peers’ loads when the replicas are updating, lead to the difficulty of replicas updating caused by replica redundancies and the low researching hit ratio for some "cold spot" files, we propose a domain index based replicas management strategy. First, we build a searching/updating model based on practical application.Then we calculate the size of a peer’s optimal domain by mathematical formula.Based on the optimal domain, so the number of replicas of P2P network is confirmed, that searching and updating information get optimization of balance. In the process of theoretical derivation we search that minimizing the number of searching and updating of the replicas based on the ratio of searching frequency and updating frequency. On that basis,we build the domain index by using the searching/updating model.Our algorithm makes peers build circle structure,and every peer organizes the domain spontaneously which can overlay with each other in order to make searching and updating get balance. And the index replicas and the file replicas are combined organically, then the strategy is adopted that updating happens in the domain and searching happens between different domains so that bandwidth can be managed better. Meanwhile, we propose an effective method which can make the domains adjust dynamically to solve the problem how to ensure the system performance when the number of peers in the network and the user actions change. Simulation results illustrate that the strategy we proposed in this paper can effectively reduce the loads of peers caused by searching and updating a replica, increase the searching rate of "cold spot" files and decrease the number of unnecessary replicas, and it can ensure the rate of success of searching higher when the peers churn.