Load balancers are software defined and distributed, which means they are
highly performant and not bottlenecked by a single appliance. They are tightly integrated with Compute Engine and GKE to allow for intelligent autoscaling based on various metrics, and health checks, so we can route traffic to the healthy instances and locations.

Load balancers are available based on geo scope (regional / global), network tier (premium / standard) and proxy or pass-through. Pass-through means client IP is preserved. Proxy means the client IP is not preserved, and a different connection is established between the LB and the backend instances.