Professional Cloud Network Engineer · Domain 1: Designing and planning a Google Cloud VPC network · 21% of the exam
Task 1.2: Designing VPC networks.
Standalone or Shared VPC and how many networks, interconnecting them with VPC Network Peering, Network Connectivity Center or Private Service Connect, IP address management, global or regional designs, MTU, and inserting network virtual appliances with routes and load balancing.
Study it
VPC design: standalone or Shared VPC, peering, Network Connectivity Center topologies and network virtual appliances
IP address planning: subnets, IPv6, BYOIP, PUPI, non-RFC 1918 ranges, internal ranges, Private NAT and MTU
Lesson coming
Sample questions
Try each one before opening the answer. Every option is explained, with the Google Cloud documentation page that proves it.
Question 1 · choose 1
A company is planning two Shared VPC networks, trusted and untrusted, both owned by the central network team. A security team's service project will run firewall appliance VMs with one network interface in each of the two networks. An architect proposes one host project per network so that each network has its own IAM bindings. What should the network engineer recommend so that the appliance VMs can be deployed?
AKeep both Shared VPC networks in one host project and attach the security team's service project to that host project
BCreate one host project per network and attach the security team's service project to both host projects
CCreate one host project per network, attach the service project to the trusted host, and peer the untrusted network to the trusted one
DCreate one host project per network and run the appliance VMs in the trusted host project with nic1 in the untrusted network
Show the answer and why
AKeep both Shared VPC networks in one host project and attach the security team's service project to that host project
Correct
A service project can attach to only one host project, so a VM that needs interfaces in both shared networks needs both networks in the same host project.
BCreate one host project per network and attach the security team's service project to both host projects
Incorrect
Each service project can be attached to only a single host project.
CCreate one host project per network, attach the service project to the trusted host, and peer the untrusted network to the trusted one
Incorrect
Peering exchanges routes; a service project VM can only place an interface in its own project's networks or in its host project's shared networks, not in a peered network.
DCreate one host project per network and run the appliance VMs in the trusted host project with nic1 in the untrusted network
Incorrect
Every interface of a multi-NIC VM in a host project must use a subnet of a Shared VPC network in that same host project.
Google's VPC design guidance recommends a single host project when resources need multiple network interfaces, because a service project attaches to only one host project. Separate host projects suit networks that need separate administration policies or exceed one project's quotas, but not a VM that spans them.
The VPC networks app1, app2 and shared-svc are VPC spokes of a Network Connectivity Center hub with the mesh topology. shared-svc is also connected by VPC Network Peering to legacy, a VPC network in a partner's organization whose subnets overlap none of the others. Workloads in app1 and app2 now need to reach legacy over internal addresses, and the partner agrees to any change on its side. Which change achieves this?
AAdd legacy to the hub as another VPC spoke and keep its peering with shared-svc in place
BTurn on export of custom routes on shared-svc's side of the peering so that the hub learns legacy's subnets
CDelete the peering between shared-svc and legacy, then connect legacy to the hub as a VPC spoke
DAdd an include export filter on the shared-svc spoke that lists legacy's subnet ranges
Show the answer and why
AAdd legacy to the hub as another VPC spoke and keep its peering with shared-svc in place
Incorrect
A VPC network that is peered with a VPC spoke can't itself be connected to the hub as a VPC spoke.
BTurn on export of custom routes on shared-svc's side of the peering so that the hub learns legacy's subnets
Incorrect
A VPC spoke can't export to the hub the peering subnet routes it imported from another network, whatever the peering's route settings.
CDelete the peering between shared-svc and legacy, then connect legacy to the hub as a VPC spoke
Correct
As a VPC spoke, legacy exports its subnet routes to the hub and imports the other spokes' routes, so app1 and app2 reach it directly.
DAdd an include export filter on the shared-svc spoke that lists legacy's subnet ranges
Incorrect
Export filters select which of the spoke's own subnet routes go to the hub; they can't make the spoke re-export routes it learned by peering.
NCC and VPC Network Peering do not chain: a VPC spoke never exports the peering subnet routes it imported, and the network on the other side of that peering is barred from joining the hub as a VPC spoke while the peering exists. Replacing the peering with a spoke connection gives every spoke full subnet route exchange with legacy.
A hub VPC network reaches a data center over Cloud Interconnect, and the data center uses 10.100.0.0/16. The hub is also peered with a partner's VPC network. The network team wants an internal range resource in the hub so that no subnet in the hub can be created in 10.100.0.0/16 and the peered network can't assign addresses from it either. Which internal range settings meet this?
AUsage FOR_VPC with peering type FOR_SELF
BUsage EXTERNAL_TO_VPC with peering type FOR_SELF
CUsage EXTERNAL_TO_VPC with peering type NOT_SHARED
DUsage FOR_VPC with peering type FOR_PEER
Show the answer and why
AUsage FOR_VPC with peering type FOR_SELF
Incorrect
FOR_VPC lets the range be associated with resources in the hub, which is what the team wants to forbid.
BUsage EXTERNAL_TO_VPC with peering type FOR_SELF
Correct
EXTERNAL_TO_VPC keeps the block from being associated with resources in the hub, and FOR_SELF stops peered networks from assigning addresses from it.
CUsage EXTERNAL_TO_VPC with peering type NOT_SHARED
Incorrect
NOT_SHARED lets peer networks use the same block, as long as they use it in a way the parent network can't see.
DUsage FOR_VPC with peering type FOR_PEER
Incorrect
FOR_PEER reserves the block for use in the peered networks, the opposite of what is wanted.
An internal range carries two independent settings. The usage type says whether the block can be associated with resources in its parent network (FOR_VPC) or is reserved for something outside it (EXTERNAL_TO_VPC). The peering type says who may use the block across VPC Network Peering: only the parent (FOR_SELF), only peers (FOR_PEER) or both (NOT_SHARED).
A network team plans a Shared VPC design in which one platform project hosts a Shared VPC network and should also use subnets of another host project's Shared VPC network as a service project. What does Shared VPC allow?
AA project is either a host project or a service project, never both at once
BA project can be both, as long as the two host projects are in the same folder
CA project can be both, but only one of its roles can be active at a time
DA project can be both, if the Shared VPC Admin role is granted on the organization
Show the answer and why
AA project is either a host project or a service project, never both at once
Correct
A project that participates in Shared VPC is either a host project or a service project.
BA project can be both, as long as the two host projects are in the same folder
Incorrect
The folder doesn't matter; the project can take only one of the two roles.
CA project can be both, but only one of its roles can be active at a time
Incorrect
Shared VPC has no switching between roles; a project takes one role.
DA project can be both, if the Shared VPC Admin role is granted on the organization
Incorrect
Organization-level admin rights don't let a project hold both roles.
In Shared VPC, a host project contains one or more Shared VPC networks, and a service project is attached to a host project to use its subnets. A participating project is one or the other, and the host and service projects must belong to the same organization.
A company short of RFC 1918 space plans a subnet that uses a privately used public IPv4 range (PUPI) that it doesn't own. Cloud Router will advertise the subnet to the data center. What should the network engineer warn about?
AGoogle Cloud requires proof of ownership before a PUPI range can be used in a subnet
BCloud Router can't advertise subnet ranges that use PUPI addresses
CThe data center might lose access to internet resources that really use those addresses
DTraffic to the PUPI subnet is billed as internet egress because the addresses are public
Show the answer and why
AGoogle Cloud requires proof of ownership before a PUPI range can be used in a subnet
Incorrect
PUPI addresses don't need to be owned to be used privately in a VPC network.
BCloud Router can't advertise subnet ranges that use PUPI addresses
Incorrect
Subnet ranges, PUPI included, can be advertised; the risk is what the receiving network can then reach.
CThe data center might lose access to internet resources that really use those addresses
Correct
A network that learns a PUPI range routes those addresses privately, so it might not reach the public resources at them.
DTraffic to the PUPI subnet is billed as internet egress because the addresses are public
Incorrect
The addresses are used privately inside the VPC network; nothing makes that traffic internet egress.
Privately used public IP addresses let a VPC network use public ranges internally without owning them. When Cloud Router advertises such a subnet to another network, that network might not be able to access internet resources at those public addresses.
A team creates a new VPC network without setting its MTU and later wants jumbo frames for a data processing cluster. What MTU does the network have, and how high can it be set?
A1,500 bytes by default, settable up to 9,000 bytes
B1,460 bytes by default, fixed for the life of the network
C1,300 bytes by default, settable up to 1,500 bytes
D1,460 bytes by default, and settable up to 8,896 bytes
Show the answer and why
A1,500 bytes by default, settable up to 9,000 bytes
Incorrect
Neither number matches: the default is 1,460 and the maximum is 8,896.
B1,460 bytes by default, fixed for the life of the network
Incorrect
A VPC network's MTU can be set to any value from 1,300 to 8,896 bytes.
C1,300 bytes by default, settable up to 1,500 bytes
Incorrect
1,300 is the lowest value you can set, not the default, and the maximum is far higher.
D1,460 bytes by default, and settable up to 8,896 bytes
Correct
VPC networks use a default MTU of 1,460 bytes, and any value from 1,300 to 8,896 bytes can be set.
VPC networks use a default MTU of 1,460 bytes. The MTU can be set to any value between 1,300 and 8,896 bytes; 1,500 (standard Ethernet) and 8,896 (the maximum) are common choices.