Native NVMe performance
Built on SPDK: user-space I/O that bypasses the kernel, with volumes served over NVMe-oF/TCP.
Synchronous replication, RF1–RF3
Every volume is synchronously replicated across 1, 2 or 3 nodes. You choose per StorageClass.
Automatic quorum-based failover
A stateless witness acts as tie-breaker and promotes a healthy replica with no split-brain.
End-to-end data integrity
Per-block CRC32C checksums, periodic scrubbing and NVMe-oF digests detect and repair silent corruption.
Encryption at rest
Per-volume AES-XTS-256, with the key held in a Kubernetes Secret. Enabled per StorageClass.
Backup and disaster recovery
Consistent snapshots, backup to S3 and continuous replication to a DR site with orderly failover and failback.