Software Productivity in DevOps
Qin Liu, Yidan Qin, Hongming Zhu, Hongfei Fan · Journal of Software · 2019
The investigation of multi-source, heterogeneous, multi-cycle data in DevOps has been attracting lots of attention in recent years.Although productivity is crucial for assuring instant release of DevOps, it has not been well studied based on merging effort features and unobserved cost features for open source software.An innovative software productivity estimation model in DevOps is proposed in this paper by recasting the definition of effort and cost.The proposed productivity model takes account of committed "outcomes" as cost instead of traditional man-month, and extended effort to consist of various commits (code, issue, scripts).Four open source projects are studied, with 95481 commits and 95828 issues in total.The experiment results illustrate the productivity changes with life cycle.The non-traditional code work ratio in productivity can represent iteration frequency of a software production and increases drastically before important releases.Thus we can monitor the life cycle and predicting large change of a production with productivity.