Match a job Paths Subjects Questions Quizzes Pricing
Intermediate Open Free

How Much Slower a Disk Seek Is Than a RAM Read

A main-memory reference takes ~100 nanoseconds. A spinning-disk seek takes ~10 milliseconds.

How many times slower is the disk seek — and what does that ratio look like at human scale?

Place an SSD random read on the same ladder.

Solution

C) 100,000× — five orders of magnitude between two things casually called "storage."

10 ms ÷ 100 ns = 10⁷ ns ÷ 10² ns = 100,000×.

At human scale: if a RAM reference took 1 second, a disk seek would take ~28 hours. That's the difference between glancing at a sticky note and mailing a letter and waiting for the reply.

The ladder (the classic "latency numbers every programmer should know"):

Operation Latency vs RAM
Main memory reference ~100 ns 1×
SSD random read ~100 µs ~1,000×
Spinning-disk seek ~10 ms ~100,000×
Same-datacenter round trip ~0.5 ms ~5,000×

This one ratio explains half of systems design: why databases obsess over keeping working sets and indexes in RAM, why one disk seek per request caps a drive at ~100 requests/second, why caches exist at every layer, and why B-trees are shaped to minimize seeks rather than comparisons.

The rule of thumb to say out loud: RAM ≈ 100 ns, SSD ≈ 100 µs, disk seek ≈ 10 ms — each storage tier is roughly 100–1,000× the one above it, so design to stay as high on the ladder as possible.

Share this question

← Back to System Design Basics practice

We use cookies for product analytics to improve OmniAtlas. See our Privacy Policy.