Optimal allocation of tasks to networked computing facilities

Abstract

Distributed allocation of computing tasks over network resources is meant to decrease the cost of centralized allocation. However, existing analytical models consider practically indistinguishable re- sources, e.g., located in the data center. With the rise of edge computing, it becomes important to account for the impact of diverse latency values imposed by edge/cloud data center locations. In this paper, we study the optimization of computing task allocation considering both the delays to reach edge/cloud data centers and the response times of servers. We explicitly evaluate the resulting performance under different scenarios. We show, through numerical analysis and real experiments, that differences in delays to reach data center locations cannot be neglected. We also study the price of anarchy of a distributed implementation of the computing task allocation and unveil important properties such as the price of anarchy being generally small, except when the system is overloaded, and its maximum can be computed with low complexity.