Q: Tesla infrastructure spans manufacturing factories, energy grids (Megapack), and vehicle fleet telemetry. How would you define SLOs that make sense across such radically different systems?

In factories, milliseconds of latency halt assembly lines, causing millions of dollars in downtime. SLOs must prioritize **Deterministic Latency & Hardware Uptime**: - **SLI**: `PLC / Industrial Robot Command Latency` (< 5ms over industrial Ethernet). - **SLO**: 99.999% of manufacturing bus commands executed under 10ms. - **Error Budget Policy**: Exceeding error budget halts automated software updates to factory lines immediately.

Energy grids manage grid stabilization and high-voltage power switching. Reliability is tied to **Electrical Safety & Grid Compliance (FERC/NERC)**: - **SLI**: Autonomous frequency regulation response time (< 200ms upon grid frequency drop). - **SLO**: 99.999% availability of autonomous grid telemetry dispatch. - **Design**: Edge autonomy—Megapacks must operate independently even during total cloud disconnect.

Managing telemetry from millions of cars over unreliable cellular networks requires prioritizing **Eventual Consistency, Data Freshness, and Ingestion Durability**: - **Availability SLI**: Ingestion pipeline success rate (`HTTP 200 / Total Ingest`) >= 99.95%. - **Freshness SLI**: Vehicle status reflected in mobile app within 3 seconds of wake-up. - **Durability SLI**: Zero data loss for safety-critical crash telemetry; non-critical battery charging logs can be locally buffered on vehicle storage.

Tie error budget consumption directly to deployment velocity. If the telemetry ingestion service exhausts its monthly budget, feature deployments freeze, and engineering capacity shifts 100% to reliability engineering.