Layer 3 · Go deeper · 7 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.
How much damage does 0.01% drift between old and new do?
Six in-process calls become six network calls. What does that cost?
Build-up · 6 steps
Strangle a booking monolith, slice by slice
An illustrative booking monolith: search 70% of requests, PNR status 15%, booking 10%, refunds 5%. Watch the route table, the ACL and the data move.
Step 1 of 6
Façade in front, all to legacy
The first release routes every request to legacy. Nothing changes for users, which proves the extra hop is safe, and the route table becomes the migration plan, made executable.
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 façade retries a failed read on legacy, hiding early bugs. Should refund writes to the new service get the same fallback?”
Your move. Pick an approach:
Deep dive 2 of 4
Interviewer
“Your booking monolith’s requests: search 70%, PNR status 15%, booking 10%, refunds 5% (illustrative). What do you carve out first?”
Your move. Pick an approach:
Deep dive 3 of 4
Interviewer
“Before its writes flip, the new refund service needs legacy refund data on every request. Where does it read it from?”
Your move. Pick an approach:
Deep dive 4 of 4
Interviewer
“Your new PNR status service is built, but nobody wrote down everything legacy does. How do you prove it before moving traffic?”
Your move. Pick an approach: