NVMe vs SSD vs HDD in server storage.
The three media types are not competing options so much as different tiers. The useful question is which part of the workload belongs on each.
The practical differences
| Medium | Strength | Typical use |
|---|---|---|
| NVMe SSD | Lowest latency, highest throughput and IOPS per drive | Databases, virtualization, AI data pipelines, scratch |
| SATA / SAS SSD | Consistent performance at lower cost per drive | General workloads, boot, mid-tier capacity |
| Nearline HDD | Lowest cost per terabyte at high capacity | Bulk capacity, backup targets, archives, media |
Endurance matters more than the interface
Two NVMe drives of the same capacity can differ several times over in write endurance. Read-intensive drives are appropriate for serving data and wrong for write-heavy logging or caching tiers, where they will wear out early.
State the write profile of the workload and let endurance class be matched to it, rather than specifying a drive model and hoping.
Capacity is not usable capacity
Protection schemes, spare capacity and filesystem overhead all sit between raw and usable. Specify usable capacity and the protection scheme together, or the quotations you receive will differ in drive count for reasons that have nothing to do with price.
Chassis constraints decide what is possible
- Backplane type determines whether NVMe is supported at all
- Bay count and form factor cap the maximum drive capacity
- Controller choice differs between SSD-only and mixed designs
- High drive counts raise rebuild times and airflow requirements
A reasonable default tiering
For most mixed environments: NVMe for the hot working set and anything latency-sensitive, SSD for general-purpose volumes, and nearline HDD for bulk retention. Where the split is genuinely unclear, state the capacity and performance target per tier and ask for the tiering to be recommended.
Describe your storage requirement.
Capacity, performance and protection, quoted against one specification.