Is there any recommendation for or relation among the server RAM quota, bucket RAM quota and index service RAM quota? For example, if I allocate 4 GB RAM to a node, how much should I allocate to index services? How do I find out what’s optimal index service RAM quota?
Good question !
Sizing can be complicated but it highly depends on your use cases and workload.
We have published some documentation regarding sizing, you can take a look here:
This is an extensive blog on 3.0 as well:
We will update the above blog and come up with recommendation for 4.0 cluster.
For your specific question, you need to look into the following considerations:
- What’s the query latency you can tolerate?
- How much index you want to be in Memory vs disk?
- What’s the index mutation rate you are expecting etc.