Opensearch Performance

Hi , Community

What is Throttling or Rate Limit of OpenSearch?

Hi. Not 100% sure what you’re after here. Did you already encounter some throttling and want to get more details? I’ve bumped into index and merge throttling, which can limit indexing performance (typically when IO can’t keep up).

Or are you interested in rate limiting, e.g. to throttle indexing so there’s headroom left for search? In this case, you may want to look at reducing thread pool sizes of what you’re trying to throttle (e.g. writes). This way, requests get queued and eventually rejected faster (clients should retry).

Yeah some more information would be helpful here. Seems like you are asking how fast OpenSearch can ingest/serve searches. The answer although not very helpful is it really depends on your use case. If you are a search heavy workload you may need fewer nodes and more replication. If you are a log collection workload then you will need to focus on having lots of primary shards and then maybe rolling into smaller replicated sets.

Testing is going to be the best way to get the answers you are looking for. Say for example you are collecting logs. Maybe take a small sample of them and see how OpenSearch performs handling them. Then you can better approximate your cluster size.