MAS-based, Scalable Allocation of Resources in Large-scale, Dynamic Environments: (Doctoral Consortium)

I study the resource allocation problem, with a focus on computing environments (Cloud, Internet of Things), by employing market mechanisms such as auctions that consider supply and demand, and dynamic pricing in the decision process. However, current approaches face two limitations: (i) centralisation, as there is typically a central entity (such as an auctioneer) that holds all the information and decides the allocation based on it, therefore either limiting the size of the auctions that can be handled, or leading to compromises in the quality of the solution, and (ii) inflexibility, or inability to deal with heterogeneous, dynamic environments. In this dissertation I propose a scalable and adaptive approach to overcome these limitations: a market-based, completely decentralised multi-agent system. I investigate techniques to enhance the performance, scalability and quality of the decentralised allocation using insights from centralised approaches.