Migrate from Elasticsearch or OpenSearch in one click.

Infino swaps Lucene for Parquet files in your bucket. Everything your apps and teams use stays the same.

Only the engine changes.

One click. Five steps.

  1. Snapshotthe cluster, to your bucket
  2. Loadeach index into Parquet
  3. Registersettings, mappings, aliases
  4. Attachevery index to OpenSearch
  5. Dashboardssaved objects imported

Moving a single index? Reindex it into an Infino index inside the same cluster.

What your team stops managing.

Elasticsearch or OpenSearch

  • Shard counts fixed at index creation
  • Replicas, allocation, rebalancing
  • Peer recovery when a node is lost
  • Disk on every node, sized for the data
  • Compute and storage scaled together

Infino

  • No shards to plan
  • One copy in object storage
  • A new node serves at once
  • A local cache, sized for speed
  • Compute and storage scaled apart

One copy instead of three.

A search cluster pays for primary shards, replica shards and snapshots. Infino keeps one copy in your bucket, and object storage keeps it safe.

Faster on the queries dashboards run.

  • 56×sum and average
  • 12×range filter with a total
  • 3.7×terms aggregation
  • 3.0×match with a scoped sum
  • 2.2×sum with a filter
  • 1.7×count with a filter

Times faster than Lucene. Median latency through OpenSearch, 10M documents per index.

Your queries keep working.

Served

  • match
  • match_phrase
  • multi_match
  • bool
  • term
  • terms
  • range
  • exists
  • prefix
  • wildcard
  • regexp
  • query_string
  • aggregations
  • knn
  • hybrid

Not yet

  • fuzzy
  • function_score
  • geo
  • suggest
  • custom analyzers

Side by side.

WhatElasticsearchOpenSearchInfino
Where data livesNode disksNode disksYour bucket
Copies you pay forThreeThreeOne
File formatLucene segmentsLucene segmentsParquet
SQLES|QLSQL pluginFull SQL with search
Hybrid searchRRFSearch pipelinesOne query, one pass
LicenseAGPL, SSPL, ELv2Apache-2.0Apache-2.0

Questions

Do I have to export from the live cluster?

No. Infino reads the snapshot in your bucket, so the old cluster can be stopped. To move one index at a time inside a running cluster, reindex it into a new Infino index.

What does the migration carry over?

Every document, plus settings, templates, pipelines, stored scripts, roles, tenants and users. Dashboards saved objects move with their own export and import.

Does this work for Elasticsearch?

Yes. Infino loads an Elasticsearch snapshot into Parquet in your bucket and serves it through OpenSearch, so clients and Kibana dashboards come across.

Do security rules still apply?

Yes. The OpenSearch security plugin decides what each user sees, and Infino applies those rules inside every query.

Where is Lucene still faster?

On sorted scans. Infino is faster on sums, averages, totals over a match or a range, and terms aggregations.

Who builds Infino?

The people who created OpenSearch.

Run the swap on your own bucket.

or run it yourself: pip install infino · Apache-2.0 · quickstart →