This research study presents the concept of DevOps (or development and operations) performance clusters to measure DevOps teams performance, based on three metrics, two agility metrics (delivery lead time and deployment frequency) and a stability metric (mean time to recover). It uses a data-driven approach to identify the performance clusters among DevOps teams and describes these clusters. It then describes how the DevOps technical capabilities (continuous integration and continuous delivery) contribute to those performance clusters using multinomial logistic regression. Finally, it uses statistical tests to describe the role of Waterfall, Scrum, Kanban development methods in the context of DevOps.Based on actual industrial data, a hierarchical clustering analysis was conducted to identify the performance clusters. The statistical analysis revealed three performance clusters existed: low, medium, and high. The multinomial logistic regression showed that higher percentages of continuous integration and delivery increase the likelihood of a higher DevOps performance cluster. Further statistical analysis revealed that the highest performance of DevOps is attained via Kanban. Additionally, the study showed that DevOps has an economic advantage of reducing the total cost of software development and delivery consistent with the Economic Order Quantity (EOQ) model. A performance-oriented capability model that is supported by a practical case study was constructed and can be used by software product organizations to manage and improve the implementation of DevOps.
|In Administrative Set: