We use Elasticsearch (managed by AWS) to search documents by running a complex query (fuzzy multi_match on many short text fields with a few additional matches). We don't change the data almost at all (max. once a year - reupload of whole data). The data is quite small (<100MB), but because of the query complexity it takes quite a lot of time to get the results (up to 1000ms). Most of the time the cluster is idle (like no requests for an hour), but when it receive requests they usually come in bulks (on avg 10 requests at once / in one second). There is only one index in the entire cluster. Do you have any recommendations for the elasticsearch configuration (of cluster/nodes/shards/index)? I was thinking about index parameters like index.search.idle.after or index.refresh_interval and the number of shards per index.
Asked
Active
Viewed 74 times
1
Beniamin H
- 2,048
- 1
- 11
- 17