RBD
Block · RBD
Ceph RBD volumes through Ceph CSI for Kubernetes PVs and PVCs. Use for databases, stateful apps, and tenant block StorageClasses.
Fusionative Storage · Fusion Ceph
Ceph you operate. Block, file, and object under one control plane.
Fusion Ceph is the first-class software-defined storage path in FusioNative: Rook deploys and manages Ceph, while the Cloud Management Software surfaces pools, StorageClasses, health, and site-level replication.
On this page
Accurate Fusion Ceph scope: Rook-operated Ceph protocols, RBD mirroring, RGW multisite, and how FusionVault stays separate.
Three protocols
Block, file, and object—standard Ceph, operated as Fusion Ceph.
RBD, CephFS, and RGW are Ceph’s native services. Fusion Ceph exposes them through Rook and Kubernetes CSI/StorageClasses from the FusioNative control plane.
CephFS
File · CephFS
Shared POSIX filesystem for ReadWriteMany workloads—datasets, shared model artifacts, and multi-pod access on the same cluster.
RGW
Object · RGW
RADOS Gateway for S3-compatible (and Swift-compatible) object storage—buckets for models, datasets, and application objects.
Control plane
Day-2 Ceph without a separate storage island.
Deploy and operate Rook Ceph beside Kubernetes, GPU, and tenancy—not as a disconnected appliance brochure.
01
Rook-operated deploy
Wizard-driven CephCluster lifecycle from the control plane: operators, OSDs, monitors, and managers without hand-rolling manifests for every site.
02
Pools & StorageClasses
Map Ceph pools to Kubernetes StorageClasses so tenants and AI projects get policy-aligned block or file volumes inside your boundary.
03
Health & capacity
See cluster health, OSD status, and pool capacity in the same operational surface as the rest of FusioNative Core—not a disconnected Ceph UI alone.
04
Estate visibility
Ceph is first-class. Other backends (for example Longhorn) can appear for lifecycle visibility; they are not claimed as Fusion Ceph replacements.
Operating path
From deploy to site-level replication.
Mirroring and multisite are optional Ceph capabilities you turn on when the estate needs peer sites—not default slogans on every install.
01
Deploy
Rook Ceph from the control plane wizard
02
Provision
Pools, CSI, and StorageClasses for tenants
03
Serve
Block, CephFS, and RGW for AI and apps
04
Replicate
RBD mirroring and RGW multisite where needed
Advanced features
Mirroring and multisite—named correctly.
RBD mirroring replicates block images between Ceph clusters. RGW multisite replicates object buckets across zones. Neither is the same as FusionVault backup.
Block replication
RBD mirroring
Replicate RBD images between Ceph clusters so block volumes can fail over or recover at a secondary site.
- Journal-based or snapshot-based modes as provided by Ceph RBD mirroring
- Asynchronous replication across clusters—plan RPO around peer lag, not instant sync marketing claims
- Fits DR for Kubernetes block volumes backed by RBD, not application-level backup by itself
Object replication
RGW multisite
Ceph RGW zone and zonegroup topology replicates object buckets across sites so S3-compatible data can live in more than one location.
- Multisite is the Ceph object path: zones, zonegroups, and realm configuration
- Bucket data and metadata follow RGW multisite rules—not the same mechanism as RBD mirroring
- Use for geo or site-level object continuity for models, datasets, and app buckets
Backup & air-gap
Where FusionVault fits
FusionVault covers backup and DR patterns (Velero/Kopia, etcd, ML models, WORM, air-gap export/import). It complements Ceph replication; it does not replace mirroring or multisite.
- Mirroring / multisite = storage-layer replication inside Ceph
- FusionVault = backup, restore plans, WORM, and air-gap workflows
- Regulated estates often need both: Ceph for SDS continuity, Vault for protected copies and offline patterns
Clear boundary
Storage replication vs backup and air-gap.
Use the right tool for the job. Fusion Ceph handles SDS and Ceph-native replication. FusionVault handles protected copies and disconnected patterns.
01
Fusion Ceph
- Rook-operated Ceph SDS
- RBD / CephFS / RGW protocols
- RBD mirroring between clusters
- RGW multisite for object zones
- StorageClasses for tenants & models
02
FusionVault
- Velero / Kopia backup paths
- etcd and ML model protection
- WORM and evidence-oriented retention
- Air-gap export / import
- Restore plans—not Ceph peer replication
Put Ceph on the same plane as your AI cloud.
Walk Rook deploy, StorageClasses, RBD mirroring, and RGW multisite with the platform team—then pair FusionVault where backup and air-gap are required.