Beta

Storage Scale

HCI storage that doesn't need its own team.

Storage Scale

The Storage Scale provides hyperconverged distributed storage across your Tophan cluster. Every node contributes its local disks to a shared pool with deduplication, compression, and erasure coding — no dedicated storage hardware required.

Features

FeatureDescriptionStatus
DeduplicationBlock-level dedup across the entire cluster. Identical data stored once regardless of which VM or application wrote it.Beta
CompressionInline compression with LZ4 (fast) or Zstd (dense). Per-volume configuration.Beta
Erasure CodingDistributed parity across nodes. Survive disk or node failures without full replication overhead.Beta
TieringAutomatic data placement across NVMe, SSD, and HDD tiers based on access patterns. Hot data on fast storage, cold data on dense storage.Planned
Thin ProvisioningAllocate 1TB, use 50GB. Space consumed on write, not on allocation.Stable
SnapshotsInstant point-in-time snapshots. Copy-on-write, space-efficient, and fast to create or rollback.Beta
ClonesInstant writable copies of any volume. Full independence from the source after clone.Beta
iSCSI TargetExport volumes as iSCSI LUNs for external consumers. CHAP authentication, multipath support.Beta
ReplicationSynchronous or asynchronous replication between clusters for disaster recovery.Planned
EncryptionAt-rest encryption with per-volume keys managed by the Vault Scale.Beta
Self-HealingAutomatic data reconstruction when disks or nodes fail. No manual intervention.Beta

How Dedup Works at Scale

Block-level deduplication operates across the entire cluster, not per-node. When any workload on any node writes a block that already exists anywhere in the cluster, no additional storage is consumed.

This has dramatic implications:

  • VM templates: 100 VMs from the same template store the base image once
  • Development environments: Identical dependencies across projects share blocks
  • Backup: Incremental backups that share blocks with the source consume almost no extra space

Combined with thin provisioning, you can safely over-allocate storage 5:1 or more for typical workloads.

Architecture

┌──────────────────────────────────┐
│        Storage Scale API         │  Volume management
├──────────────────────────────────┤
│      Distributed Metadata       │  Block map, dedup index
├─────────┬─────────┬─────────────┤
│  Node 1 │  Node 2 │   Node N    │  Local storage pools
│  NVMe   │  SSD    │   HDD       │  Tiered media
└─────────┴─────────┴─────────────┘

Every node runs a storage agent that manages local disks and participates in the distributed metadata layer. Volumes are striped across nodes for performance and protected by erasure coding for resilience. The metadata layer tracks block checksums for dedup and integrity verification.