Prioritizing On-Premises Capacity with Preferred Backends

What Preferred Backends and Automatic Capacity Draining Are

Preferred backends and automatic capacity draining are advanced traffic routing and resilience capabilities provided by Cloud Service Mesh (Anthos Service Mesh) for backend services. Preferred backends designate specific groups of endpoints—such as on-premises infrastructure represented by hybrid connectivity Network Endpoint Groups (NEGs)—to receive client requests first until their provisioned capacity is saturated. Automatic capacity draining continuously monitors the health status of individual endpoints within a backend group and automatically scales down the backend's capacity scaler to zero when health check passing rates fall below defined failure thresholds.

How It Meets the Requirements

• Cost Optimization: Configuring on-premises hybrid connectivity NEGs as preferred backends ensures that local on-premises compute capacity is fully utilized before any requests are spilled over to autoscaled Google Cloud GKE backends, minimizing cloud compute costs during baseline traffic.
• Automated Failure Protection: When the health check passing rate of a backend falls below 25%, automatic capacity draining instantly excludes that backend from global load balancing decisions, preventing cascading failures across the remaining healthy instances in that group.

Key Advantages and Capabilities

• Automated Failover and Overflow: Enables smooth overflow to Google Cloud during traffic spikes and failover during local infrastructure incidents.
• Hysteresis Recovery: Reintegrates drained backends automatically once at least 35% of endpoints remain healthy continuously for 60 seconds.
• Blast Radius Protection: Limits automatic draining so that no more than 50% of total backend service endpoints can be drained simultaneously, preventing total traffic blackholing.
• Zero-Touch Traffic Engineering: Operates dynamically at the control plane and data plane levels without requiring manual configuration updates or CI/CD script interventions.

Why This Is Optimal

This solution natively satisfies both cost-containment and high-availability criteria using built-in Cloud Service Mesh traffic policies. It eliminates the operational lag and fragility of external automation scripts while managing hybrid multi-cluster routing seamlessly.