Do I need to memorize the full latency table for interviews?

No. Memorize the bands (memory is nanoseconds, SSD and local network are microseconds, disks and cross-region are milliseconds) plus a few anchors: RAM 100 ns, SSD read 100 microseconds, same-datacenter round trip 500 microseconds, cross-region 100 to 200 ms. Everything else can be reconstructed from ratios, and interviewers care about the exponent, not the digits.

Why can't cross-region latency be engineered away?

The floor is physics. Light in fiber travels about 200 km per millisecond, and real routes are longer than great-circle distance. New York to Amsterdam is nearly 6,000 km of path, so roughly 60 ms round trip is the theoretical best, and real networks add routing and queuing on top. The only fixes are moving data closer to users or taking the hop off the synchronous path.

Are these numbers still accurate on modern hardware?

The absolute values drift; NVMe drives beat 100 microseconds and fast networks beat 10 microseconds per KB. The ratios are far more stable: memory stays about 1000x faster than durable storage, and anything leaving the datacenter stays in a different band entirely. Design decisions rest on the ratios, so the table keeps working even as the digits age.

Numbers to Knowchapter 1 of 4 · 3 lessons

Latency Reference

Orders of magnitude that justify every caching and colocation decision.

← All chapters View in cheatsheet

Two engineers are arguing about one line that reads a user's settings. Just fetch it from the database. No, cache it.

Neither of them says a number, so the argument goes round. Here are the numbers. From memory on the same machine, 100 nanoseconds. From a database in the same building, 500 microseconds, five thousand times longer. From a database on another continent, 100 milliseconds, a million times longer.

Nothing about the code changed across those three. Only how far the data sat. Every caching, colocation and batching argument you will ever sit through is that same argument, and it is settled by knowing which rung a piece of data is on.

Lessons

3 in this chapter

The Ladder from Cache to Cross-RegionNine orders of magnitude separate an L1 hit from a transpacific round trip.3 min
Ratios Beat Absolute ValuesHardware changes the digits every year; the powers of ten between rungs barely move.3 min
Spending Latency Numbers in an InterviewUse the ladder to justify caches, replicas, and async boundaries with arithmetic instead of vibes.3 min