
Classic VPN is the Cloud VPN option that supports static routing, which makes it the option that still works when the peer device can't run BGP. It offers two routing styles, route-based and policy-based, and Google recommends route-based tunnels over policy-based ones. The difference between them comes down to traffic selectors, the IP ranges each side of a tunnel declares during IPsec negotiation, and the Professional Cloud Network Engineer exam expects you to know how selectors behave in each style.
In a route-based tunnel, both the local and remote traffic selectors are 0.0.0.0/0, so the IKE negotiation never depends on specific subnet ranges. Static routes in the VPC network and on the peer decide which traffic actually enters the tunnel. When a new subnet appears on either side, you only add a route, and the tunnel keeps running without renegotiation. This keeps maintenance low and is the style Google recommends.
In a policy-based tunnel, you list the exact IP ranges on each side as traffic selectors. Every range change means updating the selectors and re-establishing the tunnel. As the list grows, there is more chance of a selector mismatch with the peer, which makes policy-based tunnels the hardest option to maintain.
Traffic still needs static routes to reach the tunnel. When you build a policy-based Classic VPN tunnel in the Google Cloud console, the console creates the matching static routes for you. When you use gcloud, you create them yourself with gcloud compute routes create, using the --next-hop-vpn-tunnel and --next-hop-vpn-tunnel-region flags.
On a Cloud VPN tunnel, the local traffic selector describes the Google Cloud side, and the remote traffic selector describes the peer network. VPC ranges belong in the local selector, and the peer's ranges belong in the remote selector. Putting the peer's ranges in the local selector reverses the sides, so the tunnel won't match the peer's proxy IDs and traffic won't pass.
Traffic selectors can enforce a security limit in the IPsec negotiation itself. With policy-based routing, the ranges you list are carried into the negotiation, so the local selector defines which VPC ranges the tunnel will carry, and a local selector of 0.0.0.0/0 offers every VPC range to the peer. A route-based tunnel leaves its selectors wide open, so with that style any restriction on reachable subnets comes from routes or firewall rules rather than from the negotiation.
An IKEv1 peer adds one more constraint. With IKEv1, each traffic selector can hold just one CIDR block, so multiple ranges on one side must be summarized into a single prefix that covers them exactly. For example, 10.10.0.0/24 and 10.10.1.0/24 summarize exactly to 10.10.0.0/23. The goal is one CIDR on each side of the tunnel that includes the intended ranges and nothing else.
In Professional Cloud Network Engineer scenarios, a peer that can't run BGP is the main reason to choose Classic VPN, and these limits explain why it isn't the default otherwise.
Because Classic VPN isn't supported over Cloud Interconnect, a design that needs IPsec encryption on top of Cloud Interconnect uses HA VPN over Cloud Interconnect instead.
Our Professional Cloud Network Engineer practice exams cover Classic VPN tunnels and traffic selectors alongside HA VPN topologies and Cloud Router BGP sessions, with detailed explanations that drill these distinctions.