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.
A 300-QPS database gets 280 QPS, plus retries. What happens after a blip?
5 million phones reconnect after an outage. How hard is the edge hit?
2,000 requests/s, and the database fails for 10 s. What does it see?
Build-up · 6 steps
Take one checkout path from 27× to 1.1×
A food-delivery checkout, app to database. First see how the retries multiply, then give each fix its own step and say what it costs.
Step 1 of 6
Three layers, three tries each
Every team added a sensible retry. Nobody owns the product, and the multipliers are zero on good days, so happy-path load tests never show it.
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
“A five-service checkout chain retries 3 times at every layer, mesh included, by default. Where should retries live?”
Your move. Pick an approach:
Deep dive 2 of 4
Interviewer
“Your client already uses exponential backoff with full jitter. A reviewer asks for a retry budget too. Do you need one?”
Your move. Pick an approach:
Deep dive 3 of 4
Interviewer
“The database failover finished 20 minutes ago, but checkout is still down and every query times out. What do you do now?”
Your move. Pick an approach:
Deep dive 4 of 4
Interviewer
“Old app versions and partner scripts retry however they like. How does the order service protect itself?”
Your move. Pick an approach: