Benchmark server
One of the major usecases for a database (#3 (closed)) was that we could then compute things offline, without doing any benchmark.
This will be possible only if a client machine (say, one of our laptops) can request new benchmarks when the results are not DB-cached. For this to be possible, the benchmark machines have to expose an API to request benchmarks, compute results and return them (or commit them to database).
An "easy" option would be to make that a REST API, since many libraries exist for that purpose, reducing the dev effort.