A year or monitoring production
The end of the year is closing fast, and I run into the following metric (below). What you can see here is one of our RavenDB production instances over the past year. We are continuously dogfooding our own software, and there is a clear indication of the results.
What you can see here is the total memory used by RavenDB (production load, fairly constant over time) for the past year. As we update RavenDB, we benefit from various optimizations, and the trend line is very encouraging.
Around August, we had a change that saved us a single allocation in some cases, here is the chance, you can see the impact it had:
We also started using a new feature in production around December, and that seems to have an additional memory cost, so we optimized that as well:
You can see the new build deployed around the 17th of the month.