TIME → How much time takes a single request? Time to get an order ready SCALABILITY LOAD → Number of requests per period of time. Number of cooks available.
👩 👨 👩 👨 👩 👨 👩 👨 👩 👨 ⁉ METRICS Traffic-based services (R.E.D.) RATE - ERROR - DURATION How many requests per second [RATE] Error percentage [ERROR] Duration/latency of each request [DURATION] OBS: Context is important (e.g. bad request)
🧾 🧾 🧾 ♂ 🧾 🧾 AUTONOMY & EVENTUAL CONSISTENCY After stopping receiving notifications all parts eventually converge to the same state. The only guarantee is that, at some point in the future, something good will happen. Data replication (when needed) Prevent long-running transactions [Saga] Compensation Actions vs Rollbacks
according to the load Failures cause a few or no effect in user experience Applications spread over hundreds/thousands of nodes RESPONSIVE experience under several conditions.