Dynamic optimization of computation resource consumption and latency for internet of things microservices in fog environment

Abstract

A modern web service is an orchestration of various loosely-coupled microservices executed in the cloud because each of them is easier to develop, manage, deploy and reuse than a large monolithic service. Use of the cloud has advantages such as ease of management and a pay-as-you-go model. In addition to these characteristics, Internet of Things (IoT) services are using dynamically connected mobile devices as either data sources or actuators. Due to its dynamic nature and low latency requirements, on-demand deployment of microservice instances (MSIs) and use of fog computing to reduce latency is necessary. In fog computing environments, distributed micro data centers (MDCs) provide computation resources, and their MDC location greatly affects the latency. In this paper, we study how to dynamically allocate computation resources for low-latency IoT services in a fog computing environment. For given MSI deployment requests, we developed a novel algorithm to dynamically optimize computation resources in the fog, whose overall latency and computation resource consumption are minimized. The optimization was defined a combination of a variant of graph centrality problem called minimum convexly-combined farness centrality (MCCFC) and variable sized dynamic bin packing (VSDBP). Our proposed algorithm is a combined algorithm of breadth-first search based MCCFC algorithm with early stopping (BFS-MCCFC-ES) and hybrid selective anyfit genetic algorithm (HSAFGA) for dynamic fog resource optimization. Through extensive simulations, we found that our algorithms provide suitable performance for fog resource optimization.