- If I plan to store X documents with Y KB in size - what's the formula for memory requirements?
- Search engines including ElasticSearch perform the best when data and index are cached in memory
- How many aggregate queries vs. filtered queries vs. scan queries would you run?
- How many request per second of each type are you planning to support?
- How often will you be taking full snapshots and incremental snapshots?
- How much memory is enough memory?
- ElasticSearch is not a database per ce, hence it 'does not look before it leaps...'
- Approach recommended by Elastic: "start out with more memory than you need and scale down to find the sweet spot"
- Do not enable memory disk swap - it is deadly for ElasticSearch performance
- Swap file might need to be disabled for other NoSql as well e.g.:Cassandra