We’ve fundamentally changed how we manage memory via swap. We’re now able to spill most memory to disk before a cluster OOM’s.
Existing customers have seen significant hydration time performance improvements since we released this change. While hydration times will be workload dependent, we have seen close to a 30% reduction across many of our customers.
Swap enables Materialize to utilize the cluster’s entire CPU and its cores during hydration. Swap has mitigated previous disk IO and CPU spikes.
Now that clusters are backed by swap, we no longer differentiate between memory and disk. They both just represent places to put bytes, which can go in all places. Going forward, users should only consider total memory utilization (memory + disk) when determining if a cluster is right sized. Stay tuned on changes to our monitoring tools in the Console & system tables which are coming soon. 👀
You can find more technical details in our blog post.
If you have any questions about how this impacts your environment, please ask Matty (via the chatbot in the right hand corner of our website) or reach out to your Account Executive to be connected with our team.