Skip to content
BENIBIX Learn

IOPS and Drive Speed

Level 3 · Drives and how fast they are

Why a drive has two speeds: how many megabytes it moves and how many small reads it finishes. What IOPS means and when it matters most.

Course contentsLesson 7 of 8

Level 2 · Graphics cards and AI chips

  1. CPU vs GPU vs TPU
  2. VRAM and Unified Memory

Level 3 · Drives and how fast they are

  1. HDD vs SSD vs NVMe
  2. IOPS and Drive Speed

Level 4 · Power and start-up

  1. Power Supply and UEFI

After this lesson you can

  • Tell IOPS from megabytes a second
  • Say which jobs need high IOPS
  • Spot an IOPS limit on a cloud disk

10 mincovers 1 glossary entries

Worth reading first: HDD vs SSD vs NVMe

Why a drive has two speeds

IOPS counts how many separate reads and writes a drive can finish in one second. A hard drive manages about 100 small, random ones. A SATA SSD manages close to 100,000. A fast NVMe drive can pass a million. This number matters more than megabytes a second for most busy work.

Megabytes a second is the speed for one big file read from start to end. A hard drive does fine at that, because the arm stays in one place. IOPS is the speed for lots of tiny pieces from all over the drive. There, the hard drive has to swing its arm every time.

A database works in tiny pieces. Each search reads a few small parts of a big file, from many places. So does a computer starting up, or a server handling many users at once. That is why moving a busy database from a hard drive to an SSD can make it many times faster.

Cloud disks often come with an IOPS limit written into the plan. A disk may allow 3,000 IOPS, even if the drive under it could do more. Once you hit the limit, every extra request waits in a queue. The disk still has space, but everything that uses it slows down.

Two numbers on a spec sheet can look alike and mean different things. A drive might say 7,000 megabytes a second and 1,000,000 IOPS. The first is for copying a film. The second is for a database with many users. A drive can be fast at one and slow at the other.

The test that matters is your own work. A tool that reads the same small, random pieces your app reads gives a real number. The figure on the box is the best case, measured in perfect conditions.

A common mistake is to buy a bigger cloud disk to fix a slow database. If the slow part is IOPS, more space does nothing, unless the plan ties IOPS to size. Check which limit you hit first, then pay for that one.

On small random reads the gap is not double or triple. It is about ten thousand times.

Worked example

A shop's database slows at lunch

An online shop runs its database on a cloud disk with a limit of 3,000 IOPS. At lunch, orders rise and the disk is asked for 5,000 a second. Pages take 4 seconds to load.

The team moves to a plan that allows 16,000 IOPS on the same size of disk. Pages load in under half a second, and the bill rises by a fifth.

The words for this lesson1

Check yourself

A busy database is slow and its cloud disk is half empty. What is the most likely limit?

Finished this lesson?

Marking it complete also marks its 1 glossary entries learned, which updates your roadmap and your saved page. 0 are already marked.