Elasticsearch vs OpenSearch
License differences.
Overview
Elasticsearch and OpenSearch are forks of the same codebase, separated by Elastic’s 2021 license change from Apache 2.0 to Elastic License plus SSPL. AWS forked at that point and maintains OpenSearch as Apache 2.0; Elastic continues developing Elasticsearch with its own license and roadmap. The choice for new clusters comes down to license requirements, ecosystem alignment (Elastic-first features like ML and ES|QL stay with Elasticsearch), and which managed offering matches the team’s cloud.
- License. Elasticsearch dual-licensed (Elastic License, AGPL since 8.x); OpenSearch Apache 2.0; some legal requirements force one or the other.
- AWS support. AWS OpenSearch Service is the managed offering; well-integrated for AWS-heavy stacks.
- Elastic ecosystem. Beats, Kibana, ES|QL stay Elastic-first; Elastic teams stay on Elasticsearch.
- Plugin ecosystem plus performance. Community plugins exist for both; performance benchmarks vary by workload, test on real data.
The approach
The practical approach is OpenSearch for AWS-heavy stacks where the managed offering reduces operational burden, Elasticsearch when Elastic-specific features (ML, premium Kibana, ES|QL) drive the requirement, OpenSearch when the permissive Apache 2.0 license is required by legal or product constraints, self-hosted either when control matters more than managed convenience, and per-cluster rationale documented so the choice survives team changes.
- OpenSearch for AWS. Managed AWS OpenSearch Service; well-integrated for AWS-native stacks.
- Elasticsearch for Elastic features. ML, premium Kibana features, ES|QL; the right choice when those features drive the requirement.
- OpenSearch for permissive license. Apache 2.0 matches some legal requirements that Elastic License or SSPL cannot.
- Self-host either plus documented choice. Both self-host with similar effort; per-cluster rationale committed for operational review.
Why this compounds
The choice compounds across the cluster lifetime. The right pick avoids the painful migration that comes with switching search engines (the data layout, query API, and operational tooling all carry over but rebuilding indexes at scale is significant work). Each correct choice produces ongoing operational fit; the team builds search-platform expertise that pays off on every new cluster.
- License clarity. Right license avoids late-stage legal review; the procurement conversation does not surface a blocking issue.
- Operational fit. Choice matches the team’s existing ecosystem; the operational surface stays consistent.
- Reduced migration risk. Right choice up front avoids the rebuild-at-scale migration; the cluster grows into the platform rather than fighting it.
- Institutional knowledge. Each cluster operated grows expertise; the team builds vocabulary that transfers across deployments.
Elasticsearch vs OpenSearch is an infrastructure discipline that pays off across years. Nova AI Ops integrates with search telemetry, surfaces cluster patterns, and supports the team’s search-platform discipline.