Google Cloud organizes its global infrastructure into two units that the Professional Cloud Architect exam expects you to understand precisely. A region is a large geographic area that contains multiple zones. A zone is an independent data center, or several data centers, within a region. The us-east4 region, for example, is made up of us-east4-a, us-east4-b, and us-east4-c. Each zone is its own physical data center cluster, and together they make up the region.
The reason this matters is fault tolerance. Zones are designed to operate independently of each other. If us-east4-a goes down, us-east4-b and us-east4-c keep running. Spreading workloads across zones is the standard pattern for keeping a system up when one data center has a problem.
Resources in the same region also communicate quickly and cost-effectively with each other. That is the other half of the design. Within a region, you get low latency between resources and you avoid egress charges that would apply if traffic crossed regions.
Anything you deploy on GCP is deployed in some zone, region, or combination of zones and regions. There is no such thing as a location-free resource. Different services, however, are scoped at different levels.
Compute Engine instances are zonal. The VM lives in one specific zone, and if that zone fails, the VM goes with it. Dataproc clusters are also zonal. The computational resources you allocate exist within one particular zone.
Cloud Storage buckets can be regional, dual-region, or multi-regional. You pick the geographic scope at creation time depending on how you want the data replicated. BigQuery datasets can be regional or multi-regional, with the same trade-off between performance and availability.
You can also assign a default region and zone at the project level so you do not have to specify them every time you create a resource.
Sometimes regions and zones basically do not matter. You spin something up in your default zone and move on. Other times they matter a lot.
The three reasons to be deliberate about placement are latency, cost, and fault tolerance. Latency goes up when your application is in one region and your users or other services are in another. Cost goes up because you can get charged for egress when data moves between regions. Fault tolerance comes from spreading resources across zones, and sometimes across regions, so a single failure does not take down the whole system.
You may also need to keep data in a specific region to comply with local regulations. That is often the deciding factor for where a dataset or bucket lives, regardless of latency or cost.
For the Professional Cloud Architect exam, you should be able to answer three things without hesitation. First, the relationship: a region contains multiple zones, and a zone is a data center or cluster of data centers within that region. Second, the scope of common services: Compute Engine and Dataproc clusters are zonal, Cloud Storage is regional or dual-region or multi-regional, BigQuery datasets are regional or multi-regional. Third, the trade-offs: zonal placement is cheap and fast within the zone but vulnerable to a zone failure, multi-region placement gives you resilience at the cost of egress and replication.
That framing is enough to get most regions-and-zones questions on the Professional Cloud Architect exam right.
My Professional Cloud Architect course covers regions and zones alongside the rest of the foundational architecture material.
