1 min lesson
Budget the milliseconds end to end
Describe what "Ask what the p99 is and where it's spent, then attack the tail" changes in practice.
Step 1 of 2
Budget the milliseconds end to end
Know where the time goes: client to gateway, admission and routing, gateway to provider, provider generation, stream back. For a sub-100ms Tab path, every hop has a number and you should be able to name which hop you'd cut first.
When asked to "make Tab faster," don't reach for a bigger machine. Ask what the p99 is and where it's spent, then attack the tail: pooled warm connections to kill handshakes, hedging at the measured p95 to cut the slow-provider tail and shedding to keep queueing delay out of the budget. Framing the answer as a latency budget you decompose, rather than "optimize the code," is the signal they're listening for.
Hedging everything inverts the goal. If you hedge at p50 instead of p95 you double your provider spend to shave a tail that wasn't slow and you can actually raise p99 by adding load. Hedging is a tail tool: it only pays when fired at the genuine slow minority.
Learn more
Optional practice