Elasticsearch vs OpenSearch: Deep Dive into Code, Features, and Licensing

Overview

Elasticsearch, built on Apache Lucene, has been a popular search and analytics engine since 2010, offering scalability for a wide range of applications. OpenSearch is a community‑driven fork created by Amazon in 2021, originally derived from Elasticsearch 7.10.2 and subsequently cleaned of non‑Apache‑licensed code.

Codebase Activity

Since the fork on 22 April 2021, the Elasticsearch repository recorded about 19,527 total commits, with 6,130 to the core server folder and 1,437 to other modules. OpenSearch logged roughly 3,727 total commits, 1,966 to its core server code, and 470 to auxiliary modules, indicating that Elasticsearch has seen roughly three times more core development activity.

Feature Comparison

Both engines provide identical basic search, analytics, and dashboard capabilities because OpenSearch originated from a mature Elasticsearch version. Differences lie mainly in Elastic’s proprietary X‑Pack extensions and features added after the fork.

Data streams API is implemented in both, though Elasticsearch recently added time‑series streams not present in OpenSearch.

Index State Management exists in both.

Both support alerting (though ElastAlert2 is recommended over built‑in solutions).

Cross‑cluster replication is available in both, but Elasticsearch’s implementation is a premium feature.

Some niche features remain exclusive to Elasticsearch, such as geoshape and geohex grid aggregations. OpenSearch’s unique capabilities are often limited to Amazon OpenSearch Service, while Elastic Cloud stays current with the latest Elasticsearch releases.

Security

Security features (authentication, RBAC, audit logging, encryption, multi‑tenancy) are part of Elastic’s X‑Pack Basic license and free for all users from version 7.0 onward, but advanced options require paid licenses. OpenSearch offers the same security stack completely free as an open‑source module.

Searchable Snapshots

Elasticsearch provides searchable snapshots as a paid feature for enterprise tiers, whereas OpenSearch’s “Ultrawarm” tier offers this capability for free. Both are offered via their respective managed services (Elastic Cloud and Amazon OpenSearch Service).

Machine Learning

Both platforms can serve ML workloads, but neither is purpose‑built for heavy AI tasks. Users can store vector fields and perform k‑NN/ANN searches, or use re‑ranking plugins like LTR. Built‑in ML solutions exist (e.g., Elastic SIEM), yet they are not intended for general‑purpose AI workloads.

Data Ingestion

After the fork, Elastic’s ecosystem (Logstash, Beats, client libraries) enforces version checks, preventing modern Logstash/Beats from connecting to OpenSearch. OpenSearch addresses this gap with Data Prepper and various connectors (Kafka, Flink, etc.).

Client Libraries

Elasticsearch enjoys a mature, extensive set of client libraries across many languages (Java, Python, .NET, JavaScript, Ruby, etc.). OpenSearch’s client ecosystem is younger, with some libraries incomplete, buggy, or lacking documentation, making direct HTTP calls sometimes more practical.

Licensing and Legal Considerations

Elasticsearch moved from the permissive Apache License to a more restrictive license that forbids offering the Elasticsearch API as a hosted service. OpenSearch remains under the Apache License. This distinction affects SaaS offerings and may introduce legal gray areas for embedding Elasticsearch in larger solutions.

Support and Documentation

OpenSearch, being community‑driven, lacks official commercial support; hosted providers (Amazon OpenSearch Service, Aiven) handle infrastructure but not usage guidance. Elastic offers paid subscriptions and Elastic Cloud support, though support scope can be limited.

Conclusion

For most standard use cases—text search, log analysis, dashboards—Elasticsearch and OpenSearch are functionally equivalent. Elasticsearch benefits from broader client support, faster bug fixes, and a larger developer community, while OpenSearch offers a fully free security stack and potentially lower hosting costs, especially when using Amazon’s managed service.