GCP TCP/UDP Load Balancer operates at Layer 4, which means it does not inspect application layer data such as HTTP headers. Therefore, application layer firewalls are not part of its architecture.
Using a global forwarding rule with multiple backend regions allows the load balancer to distribute traffic efficiently and scale globally, which is critical for handling millions of connections.
Client IP-based session affinity ensures the same client connects to the same backend, improving session persistence. However, it can cause uneven load distribution if some clients generate more traffic.
Health checks for TCP/UDP load balancers verify backend availability by checking if the backend IP and port respond, without inspecting application data.
10,000,000 connections / 50,000 per instance = 200 instances needed. Adding 20% safety margin: 200 * 1.2 = 240 instances total.