DeepSeas benchmarking system consists of salt states, a runner and this pillar data in this directory.
The file config.yml sets a few config variables that mostly govern where benchmark files are stored. This will likely grow more complex in the future
Collections accumulate a number of jobs that are to be run together. For now there is only the default collection. Ultimately the benchmark stage is to be flexible enough to run a specified collection, as well as only single jobs. This feature is not yet implemented
To add new job one needs to write a job specification and a fitting template.
Job specs can be found in their respective subdirectories (e.g. fio). The
only mandatory information in a spec file is the template variable. This
specifies which template the specification wants to use.
The template is sought in the template subdirectory. Templates are
specified in Jinja.
The benchmark runner coordinates the running of the actual benchmark jobs on the master (other minions are configured to run the actual loads). For now it only runs fio jobs on CephFS...more to come