Red Hat · Infrastruttura · YoctoIT tech page

Data Foundation

Software-defined storage inside OpenShift: block, file and object (Ceph) for the data that lives with the containers.

FOCUS · CONTAINER STORAGEKubernetes-native persistence: storage that follows the applications
Request the Free Assessment Learn more
YoctoIT material for clients and partners · Red Hat, RHEL, OpenShift, Ansible and the other products mentioned are trademarks of Red Hat, Inc. o di sue affiliate.
01 · What it is

OpenShift Data Foundation, made clear.

ODF brings Ceph inside OpenShift as native storage: block volumes for databases, shared file systems, S3 object storage — provisioned via StorageClass, replicated across nodes, managed like everything else on the platform. Data becomes a citizen of the cluster, not a guest of a SAN.

3-in-1
block, file and object from a single Ceph platform
Native
StorageClass and PVCs: storage requested in YAML, delivered in seconds
Metro-DR
synchronous replication across sites for the volumes that can't be lost
OpenShift Data Foundation
OFFICIAL RED HAT BRANDING · DATA FOUNDATION
ARCHITETTURA UFFICIALE · DATA FOUNDATION · FONTE: RED HAT DOCS
OFFICIAL ARCHITECTURE · DATA FOUNDATION · SOURCE: RED HAT DOCS
02 · How to use it well

The things that make the difference.

Inside the cluster

Applications & VMsPVCs: the storage request
Block (RBD)
File (CephFS)
Object (S3)
three interfaces, one backend
Ceph distributed across the nodesreplication and self-healing
Local NVMe disks or Fusionthe iron underneath
Storage as a cluster service

Honest sizing

Ceph wants thought-out nodes and disks: replicas, failure domains and growth are sized beforehand.

Storage for VMs

With OpenShift Virtualization, ODF gives VMs live migration and snapshots: the piece that completes the VMware exit.

Built-in resilience

Snapshots, clones and regional mirroring: data protection inside the platform.

Osservabilità

Capacity, IOPS and Ceph health in the cluster monitoring: storage is no longer a black box.

03 · In depth

Ceph inside OpenShift, done right

ODF packages Ceph (RBD for block, CephFS for file, RGW/NooBaa for object) as an OpenShift operator: ready StorageClasses, replica 3 or erasure coding, failure domains per zone. Internal mode on node disks or external to an existing Ceph; the Multicloud Object Gateway federates S3; Regional-DR and Metro-DR (stretched) give asynchronous or zero RPO to stateful workloads, VMs included.

  • 3-in-1 — RBD, CephFS, S3: three semantics from a single cluster
  • Internal/External — worker disks or external Ceph: the topology is a choice
  • Failure domain — replicas distributed by zone/rack: failure anticipated by design
  • Metro-DR — stretched cluster with arbiter: zero RPO between two nearby sites
  • NooBaa/MCG — the S3 gateway federating buckets everywhere: a single namespace
  • Capacity alert — thresholds and expansion per OSD: Ceph doesn't like 100%
04 · Numbers and lifecycle

The numbers that matter.

3x
the default replication (or EC for efficiency)
RPO 0
in stretched Metro-DR up to ~10ms RTT
S3
the internal object API: the lake inside the cluster
85%
the fill threshold never to exceed: we govern it
Container-native storage is a cluster within the cluster: we size it by failure domain and keep it far from the red thresholds.
05 · Use cases

Where it really pays off.

Databases on OpenShift

PostgreSQL, MongoDB and queues with replicated volumes.

AI & data pipelines

On-prem S3 object storage for datasets and models.

Stateful everywhere

Stateful apps without depending on an external SAN.

Containers without thought-out storage = incidents foretold: ODF sized and managed by people who operate it.