iRep: indirect reciprocity reputation based efficient content delivery in BT-like systems

Delivering content efficiently through biased neighbor selection has attracted a lot of researchers' attention in recent years. Meanwhile, how to help peers understand each other is a crucial precondition of biased neighbor selection in distributed P2P context. Recent researches have showed that reciprocation based schemes provide a feasible way of categorizing peers. These schemes can perform in direct or indirect ways. For example, tit-for-tat is a typical example of direct ways, while reputation-based schemes are the representative examples of indirect ways. Although tit-for-tat has been proved to be successful from practical deployment and academic studies, it still suffers from free-riding, malicious peers and other problems. On the other hand, reputation-based schemes pay little attention to malicious peers. However, in this paper we propose an approach to promote the whole system performance as well as cope with malicious peers. Firstly, by building a simple mathematical model for the content delivery process in BT-like system, we find that the delivery order is critical to system performance, after that we present the best selection policy from system and single peer's perspective respectively. Secondly, according to analysis results, we bring forward indirect reciprocate Reputation (iRep), detail the design issues and sketch security and overhead introduced by iRep. Finally, several simulation experiments are conducted to evaluate the performance of iRep through comparison with tit-for-tat and Eigentrust. The comparison results validate that iRep can efficiently promote system performance with limited overheads.