You are doing a lift-and-shift migration of heterogeneous on-prem backend servers served by a single network load balancer frontend. Which GCP-native deployment is appropriate?

The correct approach is to create a target pool and add all backend instances to it, then deploy this target pool behind your load balancer. A target pool is a regional resource that defines a group of instances that will receive incoming traffic from a network load balancer. This directly addresses the requirement for a single load balancer frontend serving heterogeneous backend servers. Creating a managed instance group is incorrect because the backend servers are heterogeneous, meaning they likely can't be represented by a single image for auto-scaling or auto-healing. Deploying a third-party virtual appliance adds unnecessary complexity and cost when GCP's native load balancing can handle this. Using GCP's ECMP capability with static routes is not a load balancing solution for external traffic to backend servers; it's primarily for internal routing within a VPC network.