# Apache Ozone

> Distributed object store speaking both S3 and Hadoop filesystem

Apache Ozone is an open-source object store built for analytics workloads, serving the same data through a native S3 protocol and a Hadoop-compatible filesystem interface, and scaling to billions of objects.

Website: https://ozone.apache.org

## Use it when

- An HDFS estate is being modernized in place; the same data serves both the S3 protocol and the ofs Hadoop filesystem interface, so jobs migrate gradually.
- The NameNode ceiling is the problem: namespace and block management are separated, scaling to billions of objects.
- The surrounding stack is Hadoop: Kerberos authentication, Ranger authorization, and transparent encryption at rest integrate directly.
- Erasure coding on dense commodity nodes matters for capacity economics.

## Think twice when

- You need S3 features beyond the basics; versioning, object lock, SSE through the gateway, event notifications, and lifecycle tiering are all missing or on the roadmap.
- Multi-site is required: there is no multisite replication, only snapshot diff and DistCp to another cluster.
- The team is small; Ozone Manager, SCM, DataNodes, and an S3 gateway (three OMs for HA) put its operational complexity at the high end, alongside Ceph.
- There is no Hadoop legacy at all, where MinIO or cloud object storage is simpler.

## How it runs

A self-hosted Java system from the Hadoop community: Ozone Manager holds the namespace, Storage Container Manager places data, DataNodes store it, and an S3 gateway translates. Durability is 3x Ratis replication or Reed-Solomon erasure coding per bucket.

## Details

### General

| Attribute | Value |
| --- | --- |
| License | Open source (permissive) |
| SPDX identifier | Apache-2.0 |
| Deployment | Self-hosted |
| Operational complexity | High |
| Pricing | Free |
| Language | Java |
| Repository | https://github.com/apache/ozone |
| Documentation | https://ozone.apache.org/docs/ |
| Stars | 1.3k |

### Object Storage

| Attribute | Value |
| --- | --- |
| S3 API | Compatible, via the S3 Gateway, no versioning, object lock or SSE |
| Interfaces | Object, Hadoop filesystem (ofs) |
| Egress | None: self-hosted |
| Storage tiers | None: replication or erasure coding set per bucket; lifecycle on the roadmap |
| Minimum duration | None |
| Durability model | 3x Ratis replication, or erasure coding, RS-3-2, RS-6-3, RS-10-4 |
| Multi-region | No multisite: DR by snapshot diff and DistCp to another cluster |
| Event notifications | None: S3 event notifications are on the roadmap |
| Immutability | None: object lock and versioning unsupported |
| Minimum deployment | Ozone Manager, SCM and DataNodes, plus an S3 gateway; three OMs for HA |

## Capabilities

- [Object Storage](https://matca.io/capabilities/object-storage)

---

Source: https://matca.io/apache-ozone
Last updated: 2026-08-31T11:40:47.051Z