Solution

Databases on dedicated hardware, not a shared instance.

Databases live and die by memory bandwidth and IOPS — exactly what shared cloud instances meter and cap. Run Postgres, Redis, ClickHouse, and vector stores on dedicated Compute Ultra and all-flash Molten, with managed PostgreSQL available when you want it.

The problem

The problem with cloud database hosting

Managed cloud databases meter IOPS, cap memory bandwidth on shared instances, and bill provisioned throughput by the unit. A single high-IOPS volume hits a ceiling, and reading data out incurs egress. For latency-sensitive or analytics-heavy databases, you are paying a premium to be throttled.

How Smelt solves it

Memory bandwidth as a feature

Compute Ultra delivers 0.6 to 1.7 TB/s of aggregate memory bandwidth for in-memory and analytical databases that ordinary instances starve.

Native NVMe IOPS

All-flash Molten gives databases the drives' full native IOPS at zero marginal cost — no provisioned-IOPS meter, no per-volume cap.

Managed PostgreSQL

Managed Postgres with high availability and automated backups is available as a service, running on the same dedicated hardware.

No egress on reads

Replication, backups, and analytics reads do not incur egress charges.

FAQ

Questions, answered.

Do you offer managed PostgreSQL?

Yes — managed PostgreSQL with high availability and automated backups runs on dedicated hardware as a managed service. You can also self-manage any database on bare metal.

How do you handle IOPS?

All-flash Molten nodes deliver the drives' full native IOPS — millions per node — at no marginal cost. There is no provisioned-IOPS meter or per-volume ceiling.

Is the database isolated from other tenants?

Yes. Databases run on dedicated nodes, not shared instances, so there are no noisy neighbors affecting latency.

Ready to get off the cloud meter?

Charter applications are open for the first deployment. Apply in two minutes, or join the waitlist — no payment required.