JobCache Summary report

The Highlights section shows the most important details of jobcache use: the overall cache hit rate, an estimate of the actual time saved by using jobcache compared to the same build without jobcache, and the approximate speedup relative to an uncached build. Time savings esimates are based on simulator runs of the build with and without jobcache as it performed in the build, assuming there are no conflicts and no contention for agent resources.

The Outcomes section summarizes jobcache performance by jobcache outcome: hits, misses, newslots, etc. The Jobcache outcomes graph shows this information in graphical form, with each outcome shown as a percentage of the total number of jobcached jobs. Selecting any of the jobcache status labels shows the Job Search results for jobs with that outcome. The Effective workload graph shows the portion of the total workload in the build that is cached and the portion that is not cached, taking into account the effect of jobcache on that workload. For example, if a build has two jobs each taking five minutes when cached, but jobcache enables one job to run in 1 minute, the effective cached workload is five minutes, and the effective uncached workload is five minutes—​indicating that half of the build workload was cached.

The Types section provides detailed information about jobcache usage according to each type of jobcache used in the build, such as gcc, javac, and others. Selecting any of the entries in the table shows the Job Search results for jobs matching that jobcache type and outcome. The Speedup column shows the amount that each type of jobcache accelerated jobs that used that type of jobcache. For example, a value of 10x for gcc jobcache indicates those jobs that used gcc jobcache are ten times faster than the same jobs would have been without jobcache. Note that the speedups reported for individual jobcache types may be significantly greater than the overall build speedup due to jobcache, because the overall speedup is limited by both uncached jobs and the dependency structure of the build, while the type-specific speedup simply considers the workload of the jobs in question with and without jobcache.