Estimation Cheat Sheet

Latency numbers, power-of-two table, common benchmarks. Memorize these.

Ref

Source

Jeff Dean's "Latency Numbers Every Programmer Should Know" + Alex Xu, Ch 2

Originally compiled by Jeff Dean at Google. Updated for modern hardware.

Latency Numbers (2024 Hardware)

Operation	Latency	Scale
L1 cache reference	1 ns	instant
L2 cache reference	4 ns	instant
Main memory reference (RAM)	100 ns	instant
SSD random read	16 μs	fast
Read 1 MB sequentially from memory	3 μs	fast
Read 1 MB sequentially from SSD	49 μs	fast
Read 1 MB sequentially from disk (HDD)	825 μs	okay
Round trip within same datacenter	500 μs	okay
Send packet CA → Netherlands → CA	150 ms	slow

Visual Scale

L1 cache

1ns

L2 cache

4ns

RAM

100ns

SSD random read

16μs

Datacenter round-trip

500μs

HDD sequential 1MB

825μs

Cross-continent RTT

150ms

Key takeaway

Memory is 1000× faster than SSD, which is 1000× faster than cross-continent network calls. This is why caching works. This is why CDNs exist. This is why you put compute close to data.

Power of Two Table

Power	Exact Value	Approx	Name
2¹⁰	1,024	~1 Thousand	1 KB
2²⁰	1,048,576	~1 Million	1 MB
2³⁰	1,073,741,824	~1 Billion	1 GB
2⁴⁰	1,099,511,627,776	~1 Trillion	1 TB
2⁵⁰	—	~1 Quadrillion	1 PB

Availability Table

Availability	Downtime/Year	Downtime/Day
99% (two nines)	3.65 days	14.4 min
99.9% (three nines)	8.76 hours	1.44 min
99.99% (four nines)	52.6 minutes	8.6 sec
99.999% (five nines)	5.26 minutes	864 ms

Common Estimation Formulas

QPS (Queries Per Second)

QPS = DAU × avg_queries_per_user / 86,400

Example: 100M DAU, 10 queries/user → 100M × 10 / 86,400 ≈ ~12,000 QPS

Peak QPS ≈ QPS × 2 to 5 (traffic is bursty)

Storage

Storage = daily_new_objects × avg_size × retention_days

Example: 10M new tweets/day × 300 bytes × 365 days × 5 years ≈ ~5.5 TB

Bandwidth

Bandwidth = QPS × avg_response_size

Example: 12K QPS × 1 KB response = ~12 MB/s

Number of Servers

Servers = peak_QPS / single_server_QPS

Rule of thumb: a single modern web server handles ~1,000–10,000 QPS depending on workload.

Quick Size References

What	Typical Size
A tweet (text only)	~300 bytes
A metadata record (DB row)	~1 KB
A web page	~100 KB
A compressed photo	~200 KB
A high-res photo	~2 MB
A 1-minute video (compressed)	~10 MB
86,400	seconds in a day
2.5 million	seconds in a month

🔢 Don't memorize every number. Memorize the orders of magnitude and the formulas. The interviewer cares that you can reason about scale, not that you know SSD latency to the microsecond.