Autonomic Service Placement in Fog Computing

Fog computing recently emerged as novel distributed virtualized computing paradigm, where cloud services are extended to the edge of the network, thereby increasing network capacity and reducing latencies. In fog computing, applications are composed of building blocks, called microservices, that are mapped to edge computing and communication devices, referred to as fog nodes. A crucial component in fog computing are placement algorithms that assign microservices to fog nodes, since they determine the overall system performance in terms of energy consumption, communication costs, load balancing, and others. Placement strategies for virtual machines in cloud computing abound, but are generally centralized and therefore not well suited for decentralized fog systems. In this paper, we develop a fully distributed placement strategy that jointly optimizes energy consumption of fog nodes and communication costs of applications. We follow a Markov approximation approach for the design of a fully distributed autonomic service placement strategy without central coordination or global state information. Using numerical examples, we show that our placement algorithm finds solutions that are comparable to existing centralized solutions.