Adaptive Capacity Partitioning in Cooperative Computing to Maximize Received Resources

Abstract

Users in a cooperative computing environment always have a tendency to free-ride. One can employ incentive mechanisms to prevent such behavior. Some of the cooperative computing scenarios have the same access link shared between upload and download. In such a situation, increasing upload capacity decreases the download capacity and vice versa. Optimal partitioning of link capacity between upload and download needs to be done by each user to maximize its gain (i.e., download) from the network. We model this link capacity partitioning problem as a feedback control system, where feedback (resources received) decides the number of resources to be uploaded by a user. The resulting algorithm called adaptive step size (ASZ) dynamically adjusts the partitioning of link capacity to an optimal value. To compare this approach with others, a metric 'level of optimality (U)' is introduced. U achieved by the ASZ is closer to the optimal level than the reputation-based resource allocation policy (existing scheme), thus resulting in its better performance. The ASZ is also integrated with BitTorrent, and the simulation results show that it increases the resources received by the users. The ASZ can provide an efficient solution to the problem of optimal partitioning in real-life distributed networks due to its distributed implementation, robustness to changes in network dynamics, and compatibility with the existing partitioning schemes.