An Architectural Framework for Serverless Edge Computing: Design and Emulation Tools

Abstract

We consider a Software Defined Networking (SDN)-enabled edge computing domain, where networking devices also have processing capabilities. In particular, we investigate the problem of dynamic allocation of stateless computations, that we call lambda functions, and propose an architectural framework through which requests for execution of lambda functions originated by mobile nodes can be appropriately routed to specific edge devices following a serverless model. In addition, we propose a detailed emulation environment to test the architecture. Our framework supports many possible distributed algorithms to dynamically adapt the choice where requests should be executed, in order to optimize a given performance target. In the paper we consider a few such policies, to test the flexibility of the architecture. We thus present extensive performance results of the considered policies.