Decentralized Utility Maximization for Adaptive Management of Shared Resource Pools

Abstract

Resource pools are collections of computational resources which can be shared by different applications. The goal with that is to accommodate the workload of each application, by splitting the total amount of resources in the pool among them. In this sense, utility functions have been pointed as the main tool for enabling self-optimizing behaviour in such pools. The goal with that is to allow resources from the pool to be split among applications, in a way that the best outcome is obtained. Whereas different solutions in this context exist, it has been found that none of them tackles the problem we deal with in a total decentralized way. In this paper, we then present a decentralized utility maximization approach for adaptive and optimal management of shared resource pools.