An Exploratory Case Study on Effort Estimation in Microservices

Software project management plays an important role in producing high-quality software, and effort estimation can be considered as a backbone for successful project management. Size is a very significant attribute of software by being the only input to perform early effort estimation. Even though functional size measurement methods showed successful results in effort estimation of traditional data-centric architectures such as monoliths, they were not designed for today’s architectures which are more service-based and decentralized such as microservices. In these new systems, the event concept is highly used specifically for communication among different services. By being motivated by this fact, in this study, we looked for more microservice-compatible ways of sizing microservices using events and developed a method accordingly. Then, we conducted an exploratory case study in an organization using agile methods and measured the size of 17 Product Backlog Items (PBIs) to assess how this proposed method can be useful in effort estimation in microservices. The implication from the case study is that despite performing a more accurate effort estimation using the proposed size measurement than COSMIC, we were unable to significantly outperform using the total number of events. However, our suggested approach demonstrated to us a different way to use software size in terms of events, namely, to determine the coupling complexity of the project. This finding can be beneficial specifically when evaluating the change requests.

I. INTRODUCTION

The utmost objective of all software engineering researchers and practitioners is to find ways of developing high-quality software. This objective is significantly affected by the efficiency of software project management activities. Being among the most important activities of project management, software sizing, and effort estimation have always been a challenge for the software development industry and a popular research domain in academia since 1979 [1]. It is now more and more challenging due to the fast pace of technological developments and paradigm shifts in the software development domain.