Caching test results in a large-scale continuous integration system

Continuous integration systems generate significant computational overhead, contributing to the rising energy consumption of data centers. This thesis evaluates the efficacy of test result caching to mitigate this impact utilizing a custom caching tool designed to execute tests in isolated sandboxes. The system was evaluated through a dual-phase experiment: a controlled baseline study of 557 historical commits and a live production deployment analyzing over 14 million unit test executions. The results demonstrate substantial efficiency gains, with the baseline study achieving a 91.1% reduction in CPU time and a 93.4% cache hit ratio, while in the production environment, a 71.7% reduction in CPU time was observed for computationally intensive workloads. These findings confirm that caching can significantly reduce computational resource usage whilst adding negligible overhead.