SYSTEM DESIGN ESTIMATION DRILLS

Know the numbers
senior engineers
know.

Build the engineering intuition that separates seniors from the rest. Interactive estimation drills — latency numbers, storage sizing, throughput math — so you estimate with confidence in meetings, reviews, and interviews.

5Topics·
∞Generated drills·
0Signup required
Sign in to sync your progress across devices
drill · latency-numbers
QUESTION 4 OF ∞
How long does a typical SSD sequential read take for 1 MB of data?
1_ ms
Check
Correct — within ±5%
1 MB ÷ 500 MB/s ≈ 2 ms. SSD sequential reads run 200–550 MB/s.
Scratch pad · pressure-sensitive canvas

Library

Powers of 2
The foundation of every storage and capacity calculation in system design.
2¹⁰ = 1,024 ≈ 1 K
2²⁰ = 1,048,576 ≈ 1 M
2³⁰ = 1,073,741,824 ≈ 1 G
Latency Numbers
The latency hierarchy every engineer should have memorised cold.
L1 cache ~1 ns
RAM ~100 ns
SSD ~100 µs
Network ~150 ms
Unit Conversions
Fluently move between bytes, seconds, requests/s, and bandwidth.
1 Gbps = 125 MB/s
1 day = 86,400 s
1 year ≈ 3.15 × 10⁷ s
Order of Magnitude
Quickly reason about scale — thousands, millions, billions.
10³ = thousand
10⁶ = million
10⁹ = billion
Amdahl's Law
Understand the theoretical ceiling of parallelisation gains.
Speedup = 1 / (s + p/n)
s = serial fraction
p = parallel fraction
Cloud Cost Estimation
Back-of-envelope cloud pricing — compute, memory, object storage, and egress math.
1 vCPU ≈ $20 / month
1 GB RAM ≈ $5 / month
Object storage ≈ $0.02 / GB
Egress ≈ $0.10 / GB

What you'll gain

Estimate with confidence
Each drill shows the worked formula and correct value so you internalize the reasoning — not just memorize answers.
Catch bad assumptions early
Latency hierarchy, storage sizing, throughput limits — spot the difference between 10 GB and 10 TB before it ships.
Think in orders of magnitude
Numbers stop feeling abstract. You develop a feel for scale — so back-of-envelope math feels natural in any meeting or design review.
Walk into interviews ready
Capacity estimates, latency bounds, storage sizing — ingrained, not crammed. The numbers are already in your head before you walk in.

What makes this different

Theory is widely available. xstimate is about practising until the numbers are automatic.

Type the number — don't pick it
The primary mode asks you to type your answer within ±5% tolerance. No answer bank to eliminate from — you have to actually compute the value, the exact skill interviews test.
Procedurally generated — infinite practice
Every drill is freshly generated. You can't memorise a fixed question bank, and you'll never run out of problems to practice with.
Scratch pad canvas
Work through math on a pressure-sensitive scratch pad exactly like a real whiteboard interview. No other estimation tool brings this to the browser.
Spaced repetition built in
Estimation theory is widely available. What's missing is a tool that drills the math until it's automatic. That's the gap xstimate fills.

How it works

1
Pick a topic
Choose from Powers of 2, latency numbers, unit conversions, and more. Topics unlock as you build foundations.
2
Drill with instant feedback
Estimate, submit, and see the worked solution — formula, correct value, and why it matters. Every answer builds intuition.
3
Make it stick
Spaced repetition keeps numbers in your head long-term. You'll size systems, catch mistakes, and estimate on the spot — in meetings, reviews, and interviews.
Preparing for a system design interview?
You'll walk in with the numbers already in your head — capacity estimates, latency bounds, storage sizing. No last-minute cramming, just real fluency.