Shared Filesystem

Shared Filesystem

High-speed NVMe shared filesystem — the default storage for our GPU compute, Managed Slurm and Kubernetes workloads.

Overview

All-flash NVMe storage delivering 500+ GB/s aggregate read performance. Mounted natively on Bare Metal, VM and Container instances across our Glasgow, London and Texas data centres. Managed Slurm jobs and Kubernetes pods auto-mount the filesystem at provisioning time — no manual configuration required.

Performance

Aggregate Read
500+ GB/s
Per-client IOPS
200k+
Volume Capacity
Up to multi-PB
Protocol
NFS v4.2 / SMB / S3
Mount on
Bare Metal, VM, Container

Capabilities

Auto-mounted on every form factor

Bare Metal MI350X/MI300X/H200 servers, Virtual Machines and Container pods all get the Shared Filesystem mounted at provisioning — Managed Slurm and TKS handle the mount automatically.

Optimised for AI training checkpoints

Tuned for the small-file random-read pattern of tokenizers and dataloaders, plus high-throughput sequential writes for multi-TB training checkpoints on Slurm jobs.

Snapshots & clones for hyper-parameter sweeps

Instant zero-copy clones of multi-TB datasets — perfect for parallel Slurm job arrays or Kubernetes hyper-parameter sweep workloads.

Multi-protocol access

Exposed as NFS, SMB and S3-compatible — Managed Slurm uses POSIX, TKS pods use NFS mounts, S3 clients use the S3 API.

Pricing

Performance tier (NVMe)£0.20 / GB-month
Capacity tier (HDD)£0.05 / GB-month
Snapshot retentionFree, 30 days

Ideal for

Managed Slurm distributed training checkpoint storage
Dataset staging for Bare Metal MI350X or H200 training jobs
Kubernetes inference pod model-weights mount
Bioinformatics Computing shared reference datasets (PDB, FASTA)

Ready to put Technova to work?

Talk to our team about a custom GPU cluster, managed Slurm or one of our vertical AI solutions.