Nearly every architectural decision on the exam eventually reduces to "where does this run, and what happens when that place fails?" So the geography comes first.
Regions
An AWS Region is a named geographic area — us-east-1 (N. Virginia), eu-west-1 (Ireland), ap-south-1 (Mumbai). Regions are the fundamental unit of isolation in AWS, and three properties follow from that:
- Regions are independent. A failure in one Region does not propagate to another. Nothing replicates between Regions unless you explicitly configure it.
- Data stays put. AWS does not move your data out of a Region without your instruction. This is the mechanism behind data-residency compliance.
- Most resources are Regional. A VPC, an EC2 instance, an S3 bucket, an RDS database — all live in one Region and are addressed within it.
You choose a Region on four criteria, and on the exam they are usually weighed in this order:
- Compliance and data residency. A hard constraint. "Data must remain in the EU" eliminates every non-EU Region before you consider anything else.
- Service availability. New services launch in a handful of Regions first. If the workload needs a specific service, that narrows the list.
- Latency to users. Physics. Closer is faster.
- Cost. Prices differ per Region, sometimes substantially.
Availability Zones
Each Region contains multiple Availability Zones — most modern Regions have three or more. An AZ is one or more discrete data centres with:
- independent power, cooling and physical security;
- meaningful physical separation from the other AZs (kilometres apart, on different flood plains and power grids);
- high-bandwidth, low-latency, fully redundant private links to the other AZs in the Region — typically well under a millisecond of round-trip latency.
That combination is the whole point. AZs are far enough apart to fail independently but close enough to replicate synchronously. This is why RDS Multi-AZ can keep a synchronous standby without wrecking write latency, and why the default advice for any production workload is "spread it across at least two AZs behind a load balancer".
AZ names are per-account aliases. us-east-1a in your account and us-east-1a in a colleague's account are probably different physical zones — AWS randomises the mapping so that everyone does not pile into "the first one". The stable identifier is the AZ ID (for example use1-az1), which is consistent across accounts. This matters when you are coordinating placement across accounts in an organisation.
Edge locations and the global network
Beyond Regions, AWS operates hundreds of edge locations — small points of presence in metropolitan areas worldwide. They are not places you deploy applications. They exist to terminate user connections close to the user and serve three services:
- Amazon CloudFront caches your content at the edge, so a user in Chennai fetches a cached image from a nearby edge site instead of crossing an ocean to your origin.
- Amazon Route 53 answers DNS queries from the edge, which is why DNS resolution is fast everywhere.
- AWS Global Accelerator takes traffic onto the AWS backbone at the nearest edge, then routes it privately to your Region — improving latency and jitter for traffic that cannot be cached.
Local Zones, Wavelength and Outposts
Three extensions round out the picture. Each answers "I need AWS closer than the nearest Region":
| Offering | What it is | Use it when |
|---|---|---|
| Local Zones | An extension of a Region placing compute, storage and some databases in a large metro area. | You need single-digit-millisecond latency to users in a city with no Region — media rendering, real-time gaming, live streaming. |
| Wavelength Zones | AWS infrastructure embedded inside telecom providers' 5G networks. | Mobile applications that must avoid the hop from the carrier network onto the internet — connected vehicles, AR/VR. |
| Outposts | Physical AWS racks installed in your data centre, managed by AWS, using the same APIs. | Data must legally remain on premises, or a legacy system needs microsecond-latency local access. |
Scope: the thing the exam actually tests
Knowing whether a resource is global, Regional or zonal answers a surprising number of questions on its own:
| Scope | Examples |
|---|---|
| Global | IAM users, roles and policies; Route 53; CloudFront; WAF web ACLs for CloudFront; S3 bucket namespace (names are globally unique, though the bucket itself lives in a Region) |
| Regional | VPC; S3 buckets; DynamoDB tables; SQS queues; Lambda functions; AMIs; Auto Scaling groups; ELB |
| Zonal | EC2 instances; EBS volumes; RDS instances (a Multi-AZ deployment is two zonal instances); subnets |
An EBS volume can only attach to an instance in the same AZ. If a question describes moving a volume to another AZ, the answer involves taking a snapshot (snapshots are Regional, stored in S3) and creating a new volume from it in the target AZ. If it needs to cross Regions, copy the snapshot first.