Apache Ozone
A scalable, distributed object storage system in the Hadoop ecosystem with an S3-compatible interface.
Summary
A scalable, distributed object storage system in the Hadoop ecosystem with an S3-compatible interface.
Ozone bridges the legacy Hadoop world (HDFS, YARN, MapReduce) and the modern S3-based world. It gives Hadoop-native workloads an S3 API while also supporting the Hadoop filesystem interface.
- Ozone is not a drop-in HDFS replacement. It has a different consistency model and metadata architecture (SCM + OM).
- Adoption outside the Hadoop ecosystem is limited. If you don't have legacy Hadoop workloads, MinIO or AWS S3 are more practical choices.
implementsS3 API — S3-compatible interface for Hadoop environmentssolvesLegacy Ingestion Bottlenecks — migration path from HDFSscoped_toS3, Object Storage — part of the S3-compatible ecosystem
Definition
A scalable, distributed object storage system designed for the Hadoop ecosystem, with an S3-compatible interface.
Hadoop's HDFS does not scale well beyond a single namespace. Ozone provides a Hadoop-native object store that supports both the Hadoop filesystem interface and the S3 API, bridging legacy Hadoop workloads with modern S3-based architectures.
S3-compatible storage for Hadoop-based data platforms, migration path from HDFS to object storage.
Recent developments
Apache Ozone 2.0.0 GA (August 2025) — 1,700 new features + fixes on top of 1.4. Major milestone release. The ASF positions 2.0 as the "evolution from Hadoop object store to modern cloud-native storage platform" — AI, lakehouse, and hybrid-cloud are the named target workloads. Per ASF News — Apache Ozone 2.0.0 announcement.
First object store to support Apache HBase natively. Enables low-latency read/write workloads alongside traditional object storage — Ozone collapses the "HBase needs HDFS / object store is too slow" tradeoff that historically split Hadoop clusters in two. Per Apache Ozone — 2.0.0 release notes.
S3 Gateway updated for AWS Java SDK v2 2.30.0 incompatibility. AWS SDK v2 introduced an upload-protocol change that silently appended trailer data or failed; Ozone's S3 Gateway is now updated to handle SDK v2 2.30.0 and later. Critical compatibility fix — older S3G versions silently corrupted uploads. Per Apache Ozone — 2.0.0 release notes.
AWS STS AssumeRole support added for short-lived credentials. Ozone S3G now generates temporary credentials via STS-style AssumeRole flow — eliminates long-lived access-key sprawl in multi-tenant Ozone clusters. Per Apache Ozone — 2.0.0 release notes.
S3 STANDARD_IA storage class + EC replication config support. Tiered-storage primitives now usable through the S3 API — clients can set storage class to infrequent-access at PUT time and reference erasure-coding replication configs. Closes a feature gap vs MinIO + Ceph RGW. Per Apache Ozone 2.0.0 release notes.
JDK 17 + 21 compatibility; native ARM64 builds. Ozone 2.0 builds and runs on modern Java + ARM64 — opens deployment to Graviton/Ampere clusters that the JDK-8/x86-only past locked out. Per Apache Ozone — 2.0.0 release notes and Globenewswire — ASF Announces Ozone 2.0.0.
Latest release is now 2.1.0, codename "Joshua Tree." Successor to the 2.0.0 GA covered above. Per Roadmap | Apache Ozone.
An Automatic Disk Balancer landed January 29, 2026, rebalancing data across a datanode's own disks without operator intervention. Per Blog | Apache Ozone (2026-01-30).
Core architecture: Ozone separates metadata (OM/SCM) from data storage (Datanodes). Supports RATIS 3-way replication alongside Erasure Coding, plus snapshots, quotas, TLS/TDE encryption, and Ranger ACLs. Per Overview | Apache Ozone.
At exabyte scale, Didi hit a bottleneck in Ozone Manager's HA setup — a single leader node handling all requests — and needed custom engineering (OM follower reads, heterogeneous caching) to fix latency. Tertiary source, no better coverage found. Per The Present and Future of Cloudera Object Store (Cloudera Community) (2026-06-11).
Ozone positions itself as Iceberg-ready lakehouse storage, citing POSIX compatibility, full Kerberos support, and native Hadoop/Spark/Trino integration that enables atomic commits and native locking for Iceberg tables. Per Why Apache Ozone is the Preferred Object Store for Big Data (dev.to) (2026-01-05). Sources: Apache Ozone — Release 2.0.0 release notes · ASF News — Apache Ozone 2.0.0 announcement · GitHub — apache/ozone · Cloudera — Apache Ozone Multi-Protocol Aware Storage · Globenewswire — ASF announces Ozone 2.0.0 release
Connections 4
Outbound 4
Resources 3
Official Apache Ozone project site with documentation, downloads, and community links for this Hadoop-ecosystem S3-compatible object store.
Dedicated S3 gateway documentation for Ozone, explaining how to use the S3-compatible REST interface.
Official Apache Ozone source repository including the S3 gateway, container metadata layer, and Ratis consensus integration.