Workload driven feature clustering to improve performance of a microservice architecture

Abstract

Interest in microservices architectures has increased over the last few years, with a significant increase since 2014. An increasing number of companies is evolving their architecture from a monolithic system to a microservice architecture. A microservice architecture provides more flexibility and scalability, at the cost of having a distributed system, eventual consistency and increased operational complexity. The impact of the advantages and disadvantages is defined by the size of individual microservices. Currently the size of a microservice is defined by metrics that are not related to performance and scalability. Since the large impact of the size of a microservice on the performance and scalability, metrics related to these quality attributes seem more appropriate than the existing metrics. The size of a microservice is defined by the features that it offers. As a result of these observations, this research aims to find an approach to optimize the performance of microservice architectures based on its workload. This research proposes an approach combined with accompanying proof-of-concept to alter a deployment to improve the performance. The proposed approach has been validated in a case study at AFAS, an ERP vendor in the Netherlands. This case study has validated that the approach works and has identified several interesting options for related research.