Deploy, the four environments, and the arrow everyone gets backwards
Four environments, one orthogonal axis, one architectural fact worth more marks than the other three combined, and one limitation that does the opposite of what everybody guesses.
The four environments
Cloud. Fully managed. Requires Plus or above.
Self-Hosted with control plane. You run it, control plane arrangement included. Enterprise.
Hybrid. Control plane managed for you, data plane yours. Enterprise.
Standalone server. Just the server, on its own.
The tier pattern is the easy half. The two arrangements involving a control plane you are responsible for are the two Enterprise ones. Cloud starts at Plus. Standalone is the plain option.
The axis that is not a fifth environment
Inside Cloud, and only inside Cloud, there is a second choice: serverless or dedicated. Serverless scales to zero, which is the property to hold: no idle cost, cold start on the way back up.
Orthogonal means it multiplies rather than adds. Cloud serverless and Cloud dedicated are both Cloud. And because it is Cloud-only, an option offering serverless for a self-hosted deployment describes a thing that does not exist.
The arrow, and why it points that way
Everyone gets this backwards on first read, and the reason is linguistic. We say the control plane “deploys to” the data plane, which sounds like a push, so the mental image becomes a control plane reaching in and putting a revision down. It does not. The data plane asks.
Think about what Hybrid is for: your data, your network, your compliance requirements, somebody else’s control plane. If the control plane had to initiate, you would be punching an inbound hole through your firewall for an external service. Nobody signs that off. Because the data plane polls outward, all you need is ordinary outbound access, which every corporate network already has. The connection direction is not a detail, it is the thing that makes the Enterprise story possible. Reason from that purpose and you will never get the arrow backwards under pressure.
Multi-region does the opposite of what you would guess
Four regional instances exist: GCP US, GCP EU, GCP APAC, AWS US. Four places. Now the part everybody gets wrong.
Three separate limitations, each askable on its own, so learn them as three facts rather than one vibe. The region is chosen at organization setup, once, before you have deployed anything, and it is independent of where your company is located. Four separate buildings, not one company with four offices. Your org is in a building, it cannot see the other three, and there is no removal van.
The tell: any option describing region routing, region failover, cross-region replication or a migration procedure is fabricated. Not advanced, not Enterprise-only. Fabricated.
Try it yourself
Which way does the connection go
This is an architecture question wearing a networking costume, and it is a very likely one.
Show answer
Correct answer: D — The data plane's listener polls the control plane; the control plane never initiates a connection
The control plane never initiates a connection to the data plane. The data plane runs a listener that polls outward. That is precisely what makes Hybrid viable: your data plane sits inside your own network with no inbound hole punched through the firewall for somebody else's control plane. The tempting wrong answer is the push option, because push is how most orchestration systems feel like they should work, and because we say the control plane deploys to the data plane, which sounds directional. The verb is not the network direction.
The four environments and who can have them
Name all four, and attach the plan requirement where there is one.
List the four deployment environments and note the plan requirement for each where one exists.
Reveal answer
Cloud, which requires Plus or above. Self-Hosted with a control plane, which requires Enterprise. Hybrid, which also requires Enterprise. And Standalone server. The pattern that makes the tiers memorable: the two arrangements that involve a control plane you are responsible for are the two Enterprise options, Cloud is the fully managed one starting at Plus, and Standalone is the plain server on its own.
Can one organization span two regions
You have a team in London and a team in Chicago. You want one organization covering both. What actually happens?
Show answer
Correct answer: A — An organization cannot span regions; multi-region support is not currently available
The documented position is blunt: multi-region support isn't currently available. An organization cannot span regions, cannot share billing across them, and migration between regions is unsupported. The tempting wrong answer is the routing one, because four regional instances genuinely exist, GCP US, EU and APAC plus AWS US, and the natural inference from a list of regions is that you can use more than one. Existence of instances is not the same as an organization being able to span them. The billing option is the sneakiest distractor because it concedes half the truth: billing genuinely cannot be shared, but the reason is that the org cannot span regions in the first place.
Why serverless versus dedicated is an axis, not an environment
This one trips people because it looks like a fifth environment sitting quietly at the end of the list.
What is the serverless versus dedicated distinction, which environment does it apply to, and what is the defining behaviour of serverless?
Reveal answer
It is an orthogonal, Cloud-only axis rather than a fifth environment. Within Cloud you choose serverless or dedicated. Serverless scales to zero, which is the defining property worth remembering, since it implies both a cold start and no idle cost. Because the axis is Cloud-only, any question offering serverless as a choice for a self-hosted or standalone deployment is describing something that does not exist.
Instances existing versus multi-region working
Name the inference gap that makes the region question catch people who actually know the facts.
Four regional instances exist. Why does that not mean multi-region is supported?
Reveal answer
Because the instances are four separate places an organization can live, not a distributed fabric one organization can span. Choosing the EU instance means your org lives there and only there. The list of instances answers where can I deploy; multi-region support would answer can one organization use several at once, and that answer is currently no. Reading a region list as evidence of multi-region capability is exactly the inference the question is built to catch. The region is chosen at organization setup, independently of where your company is located, and there is no migration path afterwards.