Key points are not available for this paper at this time.
There has been a fundamental shift amongst software developers and engineers in the past few years. The software development life cycle (SDLC) for a developer has increased in complexity and scale. Changes that were developed and deployed over a matter of days or weeks are now deployed in a matter of hours. Due to greater availability of compute, storage, better tooling, and the necessity to react, developers are constantly looking to increase their velocity and throughput of developing and deploying changes. Consequently, there is a great need for more intelligent and context sensitive DevOps tools and services that help developers increase their efficiency while developing and debugging. Given the vast amounts of heterogeneous data available from the SDLC, such intelligent tools and services can now be built and deployed at a large scale to help developers achieve their goals and be more productive. In this paper, we present Sankie, a scalable and general service that has been developed to assist and impact all stages of the modern SDLC. Sankie provides all the necessary infrastructure (back-end and front-end bots) to ingest data from repositories and services, train models based on the data, and eventually perform decorations or provide information to engineers to help increase the velocity and throughput of changes, bug fixes etc. This paper discusses the architecture as well as some of the key observations we have made from wide scale deployment of Sankie within Microsoft.
Kumar et al. (Wed,) studied this question.