Towards efficient usage of edge network for fog computing

Abstract

As mobile devices proliferate at network edges and fog computing brings the cloud closer to mobile devices, it is expected that increasing development and deployment of services will expedite the era of edge network. In this paper, we study the issues at network edges. First, we analyze the economics of edge network where there are complex interplays among the players such as Infrastructure provider, service provider, service user, and edge resource owner. I provide an equilibrium analysis of the market which gives an insight on how many economic benefits are obtained by each player under what condition. From the result of our analysis we propose two ap- proaches to accelerate fog computing. The first approach is designing a new MAC protocol for IoT applications in the edge network, called Bird-MAC, which is highly energy efficient in periodic data transmission applications. Second, we propose a learning framework, called SchedNet, for learning to manage network resource by schedul- ing inter-agent communications in fully-cooperative multi-agent tasks. These two approaches reduce the cost of using edge resource and improve the quality of services at network edges so that it results in additional economic benefits for all players at network edges. Consequently, this dissertation contributes to the understanding of the edge network and on how to use the edge resource efficiently by learning or designing an algorithm.