An efficient architecture for edge data center networks

Abstract

Edge AI environments are ever increasing as the amount of IoT-connected devices grow, thus rising the level of carbon emissions of such ecosystems. Therefore, sustainable AI cloud/edge systems are a must in order to minimize all those emissions, where AI plays its part in gaining efficiency. In this paper, a scheme to easily organize and optimize the computing resources in an edge data center is going to be proposed, based on a specific toroidal grid topology which minimizes distance between any pair of hosts being part of it, thus reducing energy needs. That architecture may be dynamically adjusted according to the current traffic conditions and their expected variations so as to further save power consumption by using just the necessary computing assets.