News and advice

SSD, NVMe and HDD: choose storage for your workloads

A WordPress website, a database and a collection of archived videos do not use storage in the same way. Choosing only the largest capacity or the newest acronym can therefore miss the actual requirement. Start with what your team does with its data and what is genuinely slowing that work down.

We recommend separating four questions: how much must be kept, how quickly it must be accessed, what rate of writing it needs to support and how the data will be recovered. One project can use several storage technologies, each with a distinct role.

On this page

Understand the difference between SSD and HDD

An HDD stores data on rotating magnetic platters. An SSD generally uses flash memory and has no moving read mechanism. Kingston's technical overview explains the distinction and how it influences potential uses.

An SSD is often suitable for the frequent, scattered accesses of a system or application. An HDD remains worth considering when the volume to retain weighs more heavily in the decision. This does not mean SSDs never fail or every HDD suits a NAS. Compare specific models and their compatibility with the intended equipment, rather than relying on the technology family alone.

NVMe describes communication with storage

NVMe is not a third category directly equivalent to SSD and HDD. The NVM Express specifications define how software communicates with non-volatile storage over PCI Express and other transports. An SSD can therefore use NVMe, while other SSDs use SATA.

When selecting hardware, check media, protocol, physical format and system compatibility separately. An acronym on a specification sheet does not establish that a drive fits the enclosure, is recognized by the server or works as intended. Ask for a specific configuration rather than a proposal that only says “NVMe storage.”

Describe the operations that matter

These three starting points can help frame a discussion, subject to the project's measurements and constraints:

  • Active website or application: identify when users wait, which operations occur together and what the database needs. Public website files and transactional data do not necessarily play the same role.
  • Document sharing or NAS: describe the number of people, file types, simultaneous access and growth in stored information. A proposal should also explain network access and permission management.
  • Archives or backup copies: start with the amount to retain, the agreed retention period and a recovery scenario. Storage that is inexpensive to fill may be less useful if restoring from it does not meet your needs.

These examples help formulate a request; they do not assign a technology automatically to every workload. Our guide to website, hosting-account and server backups can help define what information belongs in the scope.

Consider a practical example when comparing proposals. If employees regularly wait to open a shared project folder, describe that operation and the conditions under which it is slow. A supplier can investigate that situation more usefully than a general request to make all storage faster.

Measure more than a maximum transfer rate

Latency describes how long an operation takes; IOPS counts input/output operations per second; throughput measures the amount transferred. Kingston's testing guidance distinguishes sequential and random reads and writes, along with behaviour over sustained use.

Ask which scenario produced an advertised result. Copying one large file and serving many small simultaneous requests are different tests. Also choose a measure that your team understands, such as opening a folder, completing a processing task or loading an important page.

Storage is not always the source of a delay. A database may also need its queries and execution examined, as illustrated by the PostgreSQL performance documentation. Before changing plans, start by diagnosing hosting-resource usage. A useful recommendation should connect an observed limitation to the proposed change.

Plan for endurance and recovery

SSDs have write-endurance specifications. Read TBW and DWPD figures alongside the manufacturer's conditions and the intended workload; they do not predict an exact failure date. Kingston explains the relationship between reserved capacity, endurance and write workloads. For a write-heavy application, ask how that requirement has been considered.

Redundancy is a separate question. Some RAID arrangements can tolerate drive failures, but they do not create an independent backup: QNAP explains the distinction. Accidental deletion or the loss of the equipment calls for more than hardware continuity alone.

Before choosing a solution, clarify usable capacity, compatibility, monitoring and recovery. Our guide to planning and checking a restoration helps prepare those questions. Tell us about your workload so the storage choice follows the work it needs to support, rather than a comparison of isolated acronyms.

Keep reading