Vision of Continuously Assuring Performance

Abstract

When assuring performance, small performance differences at code level are difficult to measure at application level. Current approaches aimed at performance assurance are capable of identifying hotspots and major performance bugs. Apart from their inability to detect small regressions, they require manual effort for specification and execution. In this paper, we present the vision of continuously assuring performance by using functional unit tests. Utilizing small tests allows developers to detect small performance differences. Additionally, they do not have to define workloads manually if they use functional unit tests, since these are present in most projects. To achieve this, we propose integrating performance measurements in the continuous integration (CI) process, accelerating root cause analysis and creating parallel tests capable of identifying regressions that arise with parallel use.