Abstract

An approach to optimizing a large-scale microservice deployment for a critical notification system is presented. This paper addresses optimization for three objectives: cloud service cost, cloud resource utilization of CPU, RAM and storage, and meeting the system notification deadlines. To the best of our knowledge, this is the first work dealing with multi-objective optimization for microservice notification applications where the notification load is variable and depends on other priorly executed microservice in request processes. Furthermore, best of our knowledge, this is the first work that is based on multi-objective optimization, analyzing and proposing a Pareto front and optimal list from preselected VM candidates. First, the paper formulates multi-objective optimization problems for microservice-based large-scale notification system applications and provides the mathematical deployment model. Then, based on the expected workload and required multi-objective criteria, the paper proposes a set of algorithms that provide the Pareto front optimal solution for preselecting a Cloud Service provider’s VM type. In the end, the paper presents a case study that demonstrates the use of the proposed technique.