Hybrid Disaster Recovery Preferred Backends Automatic Capacity

What Preferred Backends and Automatic Capacity Draining Are

Cloud Service Mesh provides advanced traffic management and load balancing capabilities for hybrid and multi-region deployments. By configuring preferred backends, traffic is prioritized to a specific group of endpoints (such as on-premises compute represented by hybrid connectivity Network Endpoint Groups (NEGs)) up to their maximum capacity before spilling over to non-preferred backends. Automatic capacity draining is a resilience mechanism that automatically scales down a backend group's capacity to zero when its healthy endpoint ratio drops below a threshold.

How It Meets the Requirements

• Minimize Cloud Consumption during Normal Operations: Designating on-premises hybrid connectivity NEGs as preferred backends ensures that 100% of steady-state traffic is directed on-premises first.
• Automated Failover and Overflow: Any traffic exceeding the configured threshold or capacity of the preferred backends is automatically routed to non-preferred backends (the autoscaled Google Cloud MIGs), providing seamless failover and overflow capabilities.
• Rapid Isolation of Degraded Backends: Enabling automatic capacity draining automatically scales a backend group's capacity to zero when fewer than 25% of its endpoints pass health checks, preventing requests from being routed to failing infrastructure.

Key Advantages and Capabilities

• Zero-Touch Failover: Automatically transitions traffic to warm standby cloud instances during on-premises degradation or localized disasters.
• Cost Efficiency: Keeps active cloud resource consumption low until failover or capacity overflow is needed.
• Endpoint Protection: Protects remaining healthy instances in a degraded pool from cascading overload.
• Automatic Recovery: Automatically undrains and restores traffic to preferred backends when at least 35% of endpoints remain healthy for 60 seconds.

Why This Is Optimal

This architecture delivers an automated warm standby disaster recovery pattern without manual capacity scaling or DNS modifications, meeting both cost and business continuity objectives.