What You Need to Know to Pass the Professional Cloud Network Engineer Exam

GCP Study Hub
November 7, 2026

The Professional Cloud Network Engineer exam spans the full range of Google Cloud networking, from how a VPC network is structured to hybrid connectivity, Cloud DNS, load balancing, network security, GKE networking and troubleshooting. Questions are framed as scenarios, and the right answer usually turns on one or two details: which project owns a resource, whether a load balancer is a proxy or a passthrough design, which side of a BGP session controls a direction of traffic, or whether a service detects a threat or blocks it.

VPC network fundamentals

A VPC network in Google Cloud is a global resource, and its subnets are regional. Instances in subnets of one network reach each other across regions over internal IP addresses without gateways, tunnels or proxies, and the traffic stays on Google's backbone. A few addressing details build on that:

  • Auto and custom mode. An auto mode network creates one subnet in every region from 10.128.0.0/9, and the default network is itself auto mode. Two auto mode networks overlap, so peering between them fails. Custom mode lets you choose your own ranges.
  • Usable addresses. Google Cloud reserves four addresses in every subnet primary range, so a /24 leaves 252 usable addresses, a /27 leaves 28 and a /28 leaves just 12.
  • Ranges beyond RFC 1918. Subnets can use other supported ranges, such as the 100.64.0.0/10 shared address space.
  • Multiple interfaces. Every network interface on a VM must connect to a different VPC network, so an inline appliance between two networks needs two VPC networks.

Organizing networks across projects

When each group must be its own network administrative domain, each group gets its own VPC network. Shared VPC centralizes instead: a host project holds the VPC network, its subnets, routes and firewall rules, and service projects deploy VMs and GKE clusters into the shared subnets, with billing and IAM kept separate per project. Because the network belongs to the host project, so does anything configured on it, including the Cloud Router, which must be created in the same project as the VPC network it advertises routes for.

The predefined roles follow a split of duties. Compute Network Admin (roles/compute.networkAdmin) creates, modifies and deletes VPC networks, subnets, routes, Cloud Routers, VPN tunnels and VLAN attachments, and enabling Firewall Rules Logging on a rule takes Compute Security Admin. Google describes a Shared VPC Admin as someone who holds both Compute Shared VPC Admin and Project IAM Admin.

Routing in a VPC network

A VPC network picks the route with the most specific matching destination, and priority only breaks ties between routes with the same prefix length, with the lower number winning. The system-generated default route for 0.0.0.0/0 points to the default internet gateway at priority 1000, and it can be deleted when another default path, such as one learned from on-premises over BGP, should carry internet traffic instead. For network virtual appliances, a route whose next hop is an internal passthrough Network Load Balancer gives active-active failover with health checks. A policy-based route sends matching packets to an internal passthrough Network Load Balancer and is evaluated before subnet routes, which makes it the tool for inspecting traffic between two subnets of the same network.

Connecting VPC networks

VPC Network Peering connects two networks over internal IP addresses, across projects and organizations, with no gateway or tunnel in the path. It is not transitive, and by default it exchanges subnet routes only. Custom routes, including dynamic routes learned from on-premises, are shared only when one side exports them and the other imports them. Network Connectivity Center attaches VPC networks as spokes to a Google-managed hub that can connect a large number of them. Its preset topology decides which spokes exchange routes. In a mesh, every spoke exchanges routes with every other spoke. In a star, edge spokes exchange routes only with center spokes, so they reach shared services in the center but not each other. When a security policy requires IPsec between two VPC networks, HA VPN is Google's recommended option, and it also works across organizations.

Hybrid connectivity

HA VPN runs IPsec tunnels over an existing internet link. Dedicated Interconnect and Partner Interconnect both give private connectivity over internal IP addresses without crossing the public internet, and they differ in who supplies the physical link. Dedicated Interconnect needs your own routers in a Google colocation facility and 10 Gbps or 100 Gbps circuits. Partner Interconnect uses a supported service provider that already connects to Google, so it fits when no colocation facility is nearby or the capacity needed is small. Provisioning starts in your project, where the new VLAN attachment generates a pairing key for the provider. Cross-Cloud Interconnect connects Google Cloud to another public cloud such as AWS or Azure rather than to on-premises.

Cloud Router runs BGP for these hybrid connections and belongs to one VPC network and one region. Its BGP interfaces use link-local addresses from 169.254.0.0/16, and a Cloud Router attached to a Partner Interconnect VLAN attachment must use ASN 16550. Other values worth knowing:

  • HA VPN SLA. 99.99% requires two active tunnels, one from each of the gateway's two interfaces, with BGP running over both.
  • Tunnel throughput. About 250,000 packets per second per tunnel, roughly 1 to 3 Gbps, scaled by adding tunnels and making them all active.
  • Classic VPN. Supports static routing for peers that can't run BGP, and Google recommends route-based tunnels over policy-based ones.
  • Interconnect SLA. The 99.9% topology uses two connections in one metro, each in a different edge availability domain. The 99.99% topology uses four connections across two metros and two regions, with global dynamic routing.

Dynamic routing mode is regional by default, so learned routes reach only VMs in the Cloud Router's region. With global dynamic routing, each Cloud Router also advertises other regions' subnets with an inter-region cost of at least 201. Default advertisement mode announces only subnet routes, so anything else needs custom advertisement mode. Each direction of hybrid traffic is controlled by the side that sends the routes, lower values win, and Cloud Router carries each session's advertised route priority in the MED.

Cloud NAT and private access

Cloud NAT lets VMs with only internal addresses start outbound internet connections without allowing new inbound ones, and Public NAT only translates interfaces that have no external IP address. Each NAT IP address provides 64,512 usable source ports, and the default is 64 ports per VM. Private Google Access is enabled per subnet. private.googleapis.com (199.36.153.8/30) serves most Google APIs, while restricted.googleapis.com (199.36.153.4/30) serves only APIs that VPC Service Controls supports. Private services access reaches producer networks for services such as Cloud SQL through VPC Network Peering, while Private Service Connect exposes specific services to consumers rather than giving general IP reachability. Cloud Run services and Cloud Run functions reach a VPC network in one of two ways: Direct VPC egress, Google's recommended and cheapest option, or a Serverless VPC Access connector, whose instances are billed all the time.

Cloud DNS

Private zones answer only for authorized VPC networks, and a private zone with the same name as a public hostname gives split-horizon DNS. A forwarding zone sends one domain's queries to target name servers such as on-premises DNS, and an inbound server policy creates entry point addresses so on-premises resolvers can query Cloud DNS. A DNS peering zone resolves a domain using another VPC network's resolution order and needs no network path to it. In hub-and-spoke designs, the hub holds the hybrid DNS configuration and the spokes borrow it.

Load balancing and Cloud CDN

Application Load Balancers only understand HTTP and HTTPS. A proxy ends the client connection and opens a new one to the backend, while a passthrough Network Load Balancer delivers packets unchanged with the original client source IP, and the external passthrough Network Load Balancer is the only external load balancer that accepts UDP from internet clients. Health check probes come from 35.191.0.0/16 and 130.211.0.0/22. Cloud CDN caches at Google's edge for a global external Application Load Balancer, respects the origin's Cache-Control header and isn't supported on an external passthrough Network Load Balancer.

Network security

VPC firewall rules, part of Cloud NGFW, target instances by network tag or service account, never by resource label. Hierarchical firewall policies attach to the organization or a folder and are evaluated before VPC firewall rules and network firewall policies regardless of priority. Cloud NGFW Enterprise adds Google-managed intrusion prevention, including on decrypted TLS traffic. Google Cloud Armor policies attach to the backend services of an Application Load Balancer and, behind a global external Application Load Balancer, are enforced at Google Front Ends. VPC Service Controls protects Google-managed APIs with service perimeters, with a dry-run mode that logs violations without blocking. Cloud IDS detects threats in mirrored traffic but can't block them.

GKE networking

VPC-native clusters use alias IP ranges, with nodes in the subnet's primary range and Pods and Services in secondary ranges. Each range should be sized for the most it must ever hold, including nodes added by autoscaling, and the Service range can't be changed once the cluster uses it. In a Shared VPC, the GKE service agent needs roles in the host project, and a GKE Ingress becomes an external Application Load Balancer.

Monitoring and troubleshooting

VPC Flow Logs, enabled per subnet, record sampled flows with byte and packet counts. Firewall Rules Logging shows which rule allowed or denied a connection. Packet Mirroring copies full packets to collectors. In Network Intelligence Center, Firewall Insights reports rule hit counts, Connectivity Tests simulate a packet's path, and Network Topology keeps about six weeks of hourly snapshots.

How we would study for it

Across these areas, the Professional Cloud Network Engineer exam is mostly about distinctions: proxy versus passthrough, detection versus prevention, regional versus global, host project versus service project, and default versus custom behavior for routes and advertisements. We would study each topic by asking what requirement would make it the right answer and what would rule it out.

Our Professional Cloud Network Engineer practice exams cover the full exam outline, from VPC design and hybrid connectivity to load balancing and network security, with detailed explanations that drill these distinctions.

Get tips and updates from GCP Study Hub

arrow