Working Set Percentage Lowerbound


We have limited RAM in our cluster (3 nodes, 10 GB each). We do capacity estimation based on %20 working set percentage. We mainly use CRUD and view queries heavily (no N1QL)

But we also want to support more documents (e.g., %1 working set), assuming that the performance will be lower.

Is there a recommended lower bound for working set percentage, for our scenario (view queries, and CRUD).