Online distributed algorithm in edge-clouds

Abstract

Edge Computing is an architecture of cloud computing that focuses on low latency by bringing computation nearer to mobile devices. These mobile devices can offload tasks to a set of edge servers with low latency, enabling a set of new applications. However, by adding more variety of infrastructure, it exacerbates the challenging problem of dispatching and scheduling tasks. Traditional systems rely upon a centralized orchestrator. However, with the dynamic nature of edge networks, a centralized orchestrator might prove unreliable. Distributed consensus algorithms could be a possible solution to handle the problems caused by a centralized orchestrator. This paper presents an experimental model that attempts to solve the problem of resource allocation by adopting a distributed model to dispatch and schedule tasks. Without utilizing a centralized orchestrator, the objective is to reduce the overall weighted response time of all tasks. This paper also studies the feasibility of incorporating a distributed consensus algorithm to address some of the problems the experimental model.