Layer 3 · Go deeper · 9 min
Go deeper
Size it, build it, then make the calls an interviewer will push you on.
Back of the envelope
Size it up
Tap through each step of the arithmetic.
Four hard dependencies: how available can checkout be?
For a 99.95% checkout, how reliable must each critical dependency be?
Build-up · 5 steps
Add a replica, then count honestly
One replica, a twin, the switch in front and what they secretly share. Watch six nines on paper shrink to about three and a half.
Step 1 of 5
One replica
A single replica at 99.9% is down about 43 minutes a month, and every one of those minutes is an outage.
4 interview moments
You’re the candidate
The interviewer pushes on the hard parts. Pick an approach and see how it holds up.
Deep dive 1 of 4
Interviewer
“Your UPI-style app reports 99.94% availability, yet new users say they can’t link a bank account. How should you measure?”
Your move. Pick an approach:
Deep dive 2 of 4
Interviewer
“Payments must reach 99.99%, but your card gateway offers 99.9%. What do you do?”
Your move. Pick an approach:
Deep dive 3 of 4
Interviewer
“Your design doc claims six nines from two 99.9% replicas in one AZ, deployed together. A reviewer pushes back. What do you change?”
Your move. Pick an approach:
Deep dive 4 of 4
Interviewer
“Your grocery checkout takes ₹20 crore a month at 99.9%. Reaching 99.99% costs two engineers for a year plus a second region. Buy it?”
Your move. Pick an approach:
Build-up · 6 steps
Take one outage apart
A Swiggy-style order service has about one incident a month. Split one outage into detect, diagnose and repair, then shrink each phase.
Step 1 of 6
Up for a month, then down
A = MTBF ÷ (MTBF + MTTR). With one incident a month, MTBF ≈ 43,200 minutes. The outage is a sliver of the month, yet it sets the number.