Latency Budgets and Data Residency

1. Latency Is a Budget

A latency target is a total, and everything in the request path spends from it. Writing the budget down turns an argument about architecture into arithmetic.

target for the page                 200 ms
  DNS, already cached                 0 ms
  TLS handshake, resumed              5 ms
  edge to origin round trip          25 ms
  application work                   40 ms
  database query                     15 ms
  remaining                         115 ms