Uber Separates Scaling Intent From Execution on Kubernetes

Uber detailed its ServiceScale controller, which lets multiple orchestrators safely manage scaling of the same Kubernetes workloads. The platform runs over 100 clusters, about 4,000 services on 3 million cores and 1.5 million daily pod launches.

• ServiceScale and SSC separate scaling intent from execution in Kubernetes
• Uber platform: 100+ clusters, 4,000 services, 3M cores, 1.5M daily pods
• Read-your-own-write guardrail fixes stale informer cache issues
• Unified Failover Architecture cut provisioning from 2x to 1.3x, removing 1M+ cores