This research explores the use of resource-elastic tasks that allow servers to select the resources allowed to the task instead of the de facto method where task owners require certain resources from a server to use for the task. We refer to this idea as "elastic resource allocation". An example of this principle is the download of a task or sending results to/from a user as the amount of bandwidth allocated is proportional to the time taken.
In this work, we propose a new optimisation problem that enables resource-elastic tasks and propose three new algorithms: a scalable greedy algorithm, a critical value auction which is incentive compatible and a novel decentralised iterative auction.
This research as originally done in as an internship with Dr Sebastian Stein and in partnership with Professor Tom La Porta and Dr Fidan Mehmeti at Pennsylvania State University, with funding from ITA DIAS.