Elasticsearch Performance

Elasticsearch Performance Tips

• Use filter context — filters are cached and skip scoring (use for exact matches)
• Avoid wildcard at start — *term is very slow; use ngram analyser instead
• Size 0 for aggregations — set "size":0 when you only need aggregation results
• Use bulk API — batch inserts are 20x faster than individual indexing
• Refresh interval — set index.refresh_interval to 30s during bulk load
• disable _source compression — enable compression to reduce storage
• Shard sizing — aim for 10-50GB per shard; too many shards hurts performance

# Disable refresh during bulk indexing
PUT /myindex/_settings
{ "index": { "refresh_interval": "-1" } }

# Re-enable after load
PUT /myindex/_settings
{ "index": { "refresh_interval": "1s" } }