Online Resource Allocation Using Distributed Bidding Approaches
Military / Coalition Issue
It is important that coalition partners are able to share edge computing resources while disclosing only the information they desire. The set of algorithms we developed allow clients to disclose the utility (importance) of their jobs to server, or not, thus allowing coalition partners to keep some aspects of their jobs private. Likewise, coalition partners do not have to share information about their servers with their partners.
Core idea and key achievements
We devised and evaluated algorithms that allocate resources to users for submitted jobs. We examined various auction mechanisms that arrive at very close to optimal solutions and use centralized knowledge. We also devised simple distributed bidding algorithms which can quickly allocate resources to jobs in which the servers do not coordinate. The algorithms consider requirements for memory, bandwidth, processing, and deadlines. Our results show our algorithms are close to optimal.
Implications for Defence
Coalition partners may share edge-based computing resources without having to disclose key attributes of their jobs. This will provide a much more scalable infrastructure for performing real-time distributed analytics tasks which require high computational power. This will increase the pace at which information is available to the field.
Readiness & alternative Defence uses
The algorithms have been developed and tested in a simulation environment. More testing is required on real systems.
Resources and references
- Stein, Sebastian, Mateusz Ochal, Ioana-Adriana Moisoiu, Enrico Gerding, Raghu Ganti, Ting He, and Tom La Porta. “Strategyproof reinforcement learning for online resource allocation.” (2020): 1296-1304.
- C. Rublein, F. Mehmeti, S. Stein, T.F. La Porta, “Online Resource Allocation in Edge Computing Using Distributed Bidding Approaches”, accepted at IEEE MASS 2021.
- Bi, Fan, Sebastian Stein, Enrico Gerding, Nick Jennings, and Tom La Porta “A truthful online mechanism for allocating fog computing resources. (2019): 1829-1831.
- M. Towers, F. Mehmeti, S. Stein, T.F. La Porta, C. Rublein, G. De Mel, Auction-based Mechanisms for Resource-elastic Tasks in Edge Cloud Computing, submitted.
Penn State University, University of Southampton, IBM, IBM-UK