RECON: A novel approach to fighting unreliability in peer-to-peer grids

Peer-to-peer and grid computing are two important and successful computing paradigms. In peer-to-peer grids, resources are usually unreliable due to the lack of centralised control. Thus, it is critical to design a mechanism to enable reliable task execution in unreliable resource environments like peer-to-peer grids. In this work, we present RECON, which combines reliability and economic incentives to achieve such a goal. RECON aims at ensuring consumer tasks to be reliably executed while the economic revenue of service peers is maximised. First, we design a scheme to determine the minimum reliability threshold based on the bidding price of a consumer. Second, we propose a group of scheduling heuristics to obtain the maximum revenue of service providers under the constraint of the computed reliability threshold. Finally, we perform extensive simulations to evaluate the effectiveness of our solution.