AWS global infrastructure
Explain Regions, Availability Zones, edge locations, Local Zones and partitions, and choose a Region for a banking workload on latency, cost, features and data residency.
Assumes: .NET 10 LTS ยท AWS SDK for .NET v4 (AWSSDK.EC2 4.1.x, optional) ยท AWS CLI v2 ยท Regions us-east-1 and eu-west-2. Verified against official sources on 2026-11-17.
Learning objectives
By the end of this lesson you can:
- Explain what a Region, an Availability Zone (AZ), an edge location, a Local Zone, a Wavelength Zone and an Outpost are, using one sentence each.
- Tell the difference between an AZ name (
eu-west-2a) and an AZ ID (euw2-az1), and explain why the difference matters when two accounts or teams coordinate. - Classify a service as global, regional or zonal, and predict what happens to it when a single AZ fails.
- List the Regions and AZs your account can use, and check whether a service is offered in a Region, from the CLI and from C#.
- Choose a Region for a Harbourline Bank workload by weighing data residency, latency, service availability and price, and record the decision.
- Explain opt-in Regions and partitions (standard AWS, China, GovCloud, and the AWS European Sovereign Cloud).
Prerequisites
- What cloud computing really is and Public, private and hybrid cloud
- Secure your AWS account โ the Region guardrail SCP from Module 0 affects the commands in this lesson
- Tools: AWS CLI v2 with the
learn-sandboxSSO profile signed in (aws sso login --profile learn-sandbox)
Why this matters in banking
Harbourline Bank's risk committee asks three questions about any new AWS workload before it reaches design review:
- Where physically will customer data live? A UK retail bank's data-protection officer, its regulators and its customers' contracts all care about the answer.
- What happens if a data centre floods? The bank's operational-resilience rules require it to stay within an "impact tolerance" for important business services, such as making payments.
- Why does this cost more in London than in Virginia? Finance wants to know whether the Region choice is justified.
All three are answered by understanding AWS's physical layout. A developer who deploys "wherever the console happened to be pointing" can put customer statements in the wrong country, build a single-data-centre system that a regulator would reject, or pay for cross-Region data transfer nobody planned. This lesson gives you the vocabulary and the habits to avoid all three.
The idea in plain language
AWS runs data centres all over the world. It groups them in a strict hierarchy:
- A Region is a geographic area (for example London,
eu-west-2) containing several separate groups of data centres. Each Region is independent: its services, its endpoints and, by default, your data stay inside it. You choose the Region every time you create something. - An Availability Zone (AZ) is one of those groups inside a Region: one or more data centres with their own power, cooling and networking, far enough from the other AZs that a fire, flood or power cut is unlikely to hit two at once, but close enough (connected by fast private fibre) that you can replicate data between them synchronously. Every Region has at least three AZs.
- An edge location (a CloudFront point of presence) is a smaller site in a city, used to cache content and terminate connections close to users. You do not run your servers there.
- A Local Zone is an extension of a Region placed in a big city that has no Region of its own, so latency-sensitive workloads can run a few milliseconds from users.
- A Wavelength Zone puts AWS compute inside a mobile operator's 5G network; an Outpost is AWS hardware installed in your data centre.
Analogy: Think of Harbourline Bank's physical estate. A Region is a country where the bank has a licence. Availability Zones are separate operations centres in that country, each with its own generator and its own building, connected by private fibre โ if one is evacuated the others keep processing payments. Edge locations are cash machines in every high street: they serve customers quickly but do not hold the ledger. A Local Zone is a satellite office in a city without its own operations centre, and an Outpost is a rack of head-office equipment installed in a partner's building.
What problem the AWS service solves
The global infrastructure is not a service you switch on; it is the shape of everything you deploy. It exists to solve four problems that on-premises banks pay a lot to solve themselves:
| Problem | How the infrastructure solves it | Honest limitations |
|---|---|---|
| A single data centre is a single point of failure | Multiple AZs per Region, with low-latency links so you can spread one application across them | You must deploy across AZs. A single EC2 instance or a single-AZ database is still a single point of failure |
| Customers are far away | Regions on several continents, plus 750+ CloudFront points of presence and 15 Regional edge caches (CloudFront features page, 2026-11-17) | Edge locations cache and route; they do not run your database |
| Laws and contracts require data to stay in a country | You pick the Region; AWS states it will not move or replicate your content out of the Regions you choose without your agreement (except where needed to provide a service you started, or to obey the law) | AWS staff access, support cases, logs you ship elsewhere and global services are separate questions. Data residency is your design responsibility |
| Disasters affect whole areas | Regions are isolated from each other, so a second Region can be a recovery site | Nothing is replicated across Regions unless you configure it โ and you pay for the transfer |
Numbers on 2026-11-17 (they change every few months, so always check the official global infrastructure page):
- 39 Regions and 124 Availability Zones worldwide, with 2 more Regions (Saudi Arabia and Chile) announced.
- Of those, 34 Regions are reachable from a normal ("standard") AWS account. The rest sit in separate partitions: two AWS GovCloud (US) Regions, two China Regions, and the AWS European Sovereign Cloud Region
eusc-de-east-1in Brandenburg, Germany, which became generally available in January 2026. - 33 Wavelength Zones and Local Zones in 30+ metropolitan areas.
Partitions: worlds that do not touch
A partition is a completely separate copy of AWS with its own accounts, IAM and endpoints. The standard partition is aws. Others include aws-cn (China), aws-us-gov (GovCloud) and aws-eusc (European Sovereign Cloud). An account belongs to exactly one partition, and you cannot see or reach another partition's Regions from it. You can spot the partition in every ARN: arn:aws:s3:::my-bucket versus arn:aws-eusc:....
The European Sovereign Cloud matters to banks with EU operations: AWS describes it as physically and logically separate from other Regions, with its own IAM and billing systems, operated by EU residents located in the EU, and contracted through an EU entity. It is not "eu-central-1 with extra paperwork" โ moving a workload there means new accounts, new identities and a check that every service you need exists there.
Opt-in Regions
Regions launched after 20 March 2019 are disabled by default. You must enable them for each account before you can create anything there. Of the European Regions, Frankfurt, Ireland, London, Paris and Stockholm are enabled by default; Milan, Spain and Zurich must be opted in. This is a useful safety feature: an attacker with stolen credentials cannot quietly start crypto-mining in a Region you never enabled.
Microsoft / Azure comparison
| You know (Microsoft/Azure) | AWS equivalent | What's different |
|---|---|---|
Azure region (uksouth, westeurope) | Region (eu-west-2, eu-west-1) | Same idea. AWS Region codes are <area>-<direction>-<number> and you pass them to every SDK client |
| Azure availability zones (zone 1, 2, 3) | Availability Zones (eu-west-2a/b/c) | Every AWS Region has at least three AZs; not every Azure region has zones. AWS also exposes stable AZ IDs |
| Logical zone numbers that map to different physical zones per subscription | AZ names mapped independently per account in older Regions | Same trap. Use AZ IDs (euw2-az1) when comparing across accounts. Accounts created from November 2025 get a consistent mapping |
| Azure region pairs | No automatic pairing | AWS does not pick a partner Region for you; you design cross-Region recovery yourself |
| Azure Front Door / Azure CDN points of presence | CloudFront edge locations and Regional edge caches | CloudFront is a global service you configure in us-east-1 |
| Azure Extended Zones | Local Zones | Local Zones hang off a "parent" Region and support a subset of services |
| Azure Stack Hub / Azure Local | AWS Outposts | AWS-owned hardware in your building, managed from the parent Region |
| Azure Government, Azure China (21Vianet) | AWS GovCloud (US), AWS China Regions | Separate partitions with separate accounts |
| EU Data Boundary / Microsoft sovereign cloud offerings | AWS European Sovereign Cloud | A separate partition, not a configuration switch |
| Global resources in Azure (Entra ID, Front Door) | Global services (IAM, Organizations, Route 53, CloudFront) | Their control planes live in us-east-1, which matters for SCPs and outages |
Architecture
The hierarchy, from widest to narrowest:
Harbourline Bank's target layout for a customer-facing statements service โ a UK bank with an EU subsidiary:
Security-boundary view:
- The Region is the data-residency boundary. Customer content stays where you put it unless you (or a service you configure) copy it out.
- The AZ is the failure boundary. Design so that losing one AZ loses capacity, not data or availability.
- The edge is the exposure boundary. CloudFront and AWS WAF (Module 5) are where internet traffic first touches your system; cached content there must never include personal data unless you explicitly design for it.
- Global services are a control boundary. IAM changes made anywhere take effect everywhere, so they are protected by Identity Center, SCPs and logging rather than by Region.
Global, regional and zonal services
| Scope | Meaning | Examples | What an AZ failure does |
|---|---|---|---|
| Global | One control plane for the whole partition | IAM, AWS Organizations, Route 53, CloudFront, Billing | Nothing directly; these are designed for high availability by AWS |
| Regional | You choose a Region; AWS spreads the service across that Region's AZs for you | S3 buckets, Lambda, DynamoDB, SQS, SNS, KMS keys, VPCs | Usually nothing visible โ AWS handles AZ redundancy inside the service |
| Zonal | The resource lives in one AZ you pick | EC2 instances, EBS volumes, subnets, a single-AZ RDS instance | The resource is unavailable until the AZ recovers. You must run copies in other AZs |
Two sharp edges for .NET developers: an S3 bucket name is unique across the whole partition, but the bucket and its objects live in one Region; and a VPC is regional while each subnet sits in exactly one AZ (Module 2).
Step-by-step implementation
You will survey the infrastructure your sandbox account can see, then make a recorded Region decision. Nothing is created, so it costs nothing.
Sign in and confirm where you are
aws sso login --profile learn-sandbox
aws configure get region --profile learn-sandboxThe second command prints the default Region stored in your profile (expected: us-east-1). Every command below that omits --region uses this value.
List Regions and their opt-in status
Use the describe-regions command from the AWS CLI section with --all-regions. You will see the 34 standard-partition Regions, each marked opt-in-not-required, opted-in or not-opted-in. Notice that GovCloud, China and European Sovereign Cloud Regions are absent: they are in other partitions.
List the Availability Zones in us-east-1 and eu-west-2 with their IDs
Run describe-availability-zones for both Regions. Write down the ZoneName โ ZoneId pairs. us-east-1 has six AZs; London has four AZ IDs in the documentation, but your account may see three โ AWS hides constrained AZs from newer accounts.
Try a Region your guardrail blocks
Run the same command against eu-central-1. Because of the Module 0 SCP, you should get an access-denied error that mentions an explicit deny in a service control policy. That is the guardrail proving that Region choice is an access-control decision, not just a dropdown.
Check service availability before you design
Use the Systems Manager public parameters (shown in the CLI section) to list the Regions where a service is offered โ for example, confirm that Lambda and DynamoDB exist in eu-west-2. For newer features, also check the service's own documentation: a service being "in" a Region does not mean every feature is.
Compare price
Prices differ by Region. From the AWS Price List API (published 2026-11-16), S3 Standard storage for the first 50 TB per month costs:
| Region | USD per GB-month |
|---|---|
| us-east-1 (N. Virginia) | 0.023 |
| eu-west-1 (Ireland) | 0.023 |
| eu-west-2 (London) | 0.024 |
| eu-central-1 (Frankfurt) | 0.0245 |
For 20 TB of archived statements (20,480 GB): London โ $491.52/month, N. Virginia โ $471.04/month. The ~$20 difference is real but tiny next to the regulatory risk of storing UK customer data in the US without a lawful basis. Compute, data transfer and some managed services show bigger regional differences; always price your actual workload in the AWS Pricing Calculator.
Record the decision
Write a short architecture decision record (ADR) using the template in the Banking use case. Banks expect Region choices to be documented, reviewed and revisited.
AWS Console instructions
All navigation checked against the official docs on 2026-11-17; menu wording can shift slightly.
- Region selector: top-right of every console page, next to your account name. Global services (IAM, Organizations, Route 53, CloudFront) show Global instead of a Region.
- See every Region and AZ with IDs: open AWS Global View (search "Global View" in the console search bar) โ Regions and Zones โ Regions tab โ select a Region โ View details โ Zones tab. The Zone ID column shows the AZ ID.
- Enable an opt-in Region: AWS Global View โ Regions and Zones โ select the Region โ Enable Region. Alternatively account menu โ Account โ AWS Regions. In an Organization, the management account can also enable Regions for member accounts.
- Check where resources exist: AWS Global View also shows counts of common resources per Region โ a quick way to spot something left running in the wrong Region.
Opt-in Regions and the Region guardrail
Enabling a Region and allowing it in your SCP are two separate controls. Your sandbox SCP denies everything outside us-east-1 and eu-west-2, so even if you enable Milan, you could not use it. Keep both lists aligned and short.
AWS CLI
Commands are written for PowerShell on Windows and use the learn-sandbox profile.
# 1. All Regions in the standard partition, with opt-in status, as a table.
# --all-regions includes Regions you have not enabled.
aws ec2 describe-regions --all-regions `
--query "Regions[].{Region:RegionName,OptIn:OptInStatus}" `
--output table --profile learn-sandbox
# 2. Regions enabled for this account (default + opted in), via the Account API.
aws account list-regions --region-opt-status-contains ENABLED_BY_DEFAULT ENABLED `
--query "Regions[].RegionName" --profile learn-sandbox
# 3. Availability Zones in us-east-1: name, stable ID and state.
# The filter hides Local Zones and Wavelength Zones.
aws ec2 describe-availability-zones --region us-east-1 `
--filters Name=zone-type,Values=availability-zone `
--query "AvailabilityZones[].{Name:ZoneName,Id:ZoneId,State:State}" `
--output table --profile learn-sandbox
# 4. The same for London.
aws ec2 describe-availability-zones --region eu-west-2 `
--filters Name=zone-type,Values=availability-zone `
--query "AvailabilityZones[].{Name:ZoneName,Id:ZoneId}" `
--output table --profile learn-sandbox
# 5. Expect AccessDenied (explicit deny in an SCP) โ the Module 0 guardrail.
aws ec2 describe-availability-zones --region eu-central-1 --profile learn-sandbox
# 6. Include Local Zones and Wavelength Zones that exist in a Region (opted-in or not).
aws ec2 describe-availability-zones --region us-east-1 --all-availability-zones `
--query "AvailabilityZones[?ZoneType!='availability-zone'].{Name:ZoneName,Type:ZoneType,OptIn:OptInStatus}" `
--output table --profile learn-sandbox
# 7. Which Regions offer AWS Lambda? (AWS-owned public parameters, free to read.)
aws ssm get-parameters-by-path --region us-east-1 `
--path /aws/service/global-infrastructure/services/lambda/regions `
--query "Parameters[].Value" --output text --profile learn-sandbox
# 8. Long name of a Region code (prints "Europe (London)").
aws ssm get-parameter --region us-east-1 `
--name /aws/service/global-infrastructure/regions/eu-west-2/longName `
--query "Parameter.Value" --output text --profile learn-sandboxWhat each one teaches:
describe-regionsis an EC2 API, but it describes the account's Regions; the SCP exempts it so it works from anywhere.account list-regionsis the newer Account Management API that also covers opt-in status.- and 4. The
ZoneIdcolumn is the stable identifier; compare it with a colleague's output to see whether yourus-east-1ais the same building as theirs. - Region choice is enforced by access control. Read the error: it names the SCP as the reason.
- Local Zones appear with
ZoneTypelocal-zone; you must opt in to each zone group before using it. - and 8. AWS publishes infrastructure metadata as public Systems Manager parameters that any account can read. Query a Region that supports these paths (us-east-1 does); a few newer Regions do not, as the documentation notes.
Infrastructure as Code
You rarely hard-code AZ names in IaC. Ask the platform which AZs exist and pick by ID when two accounts must line up (for example, a shared network and a workload account).
terraform {
required_providers {
aws = {
source = "hashicorp/aws"
version = "~> 6.0"
}
}
}
provider "aws" {
region = "eu-west-2"
profile = "learn-sandbox"
}
# Only real AZs that are currently available (no Local or Wavelength Zones).
data "aws_availability_zones" "available" {
state = "available"
filter {
name = "zone-type"
values = ["availability-zone"]
}
}
output "az_names" {
value = data.aws_availability_zones.available.names
}
# Stable IDs: use these when coordinating with another account.
output "az_ids" {
value = data.aws_availability_zones.available.zone_ids
}terraform init then terraform plan prints the outputs without creating anything (data sources only read).
Complete code
This optional console app lists the AZs your profile can see in a Region, with names and IDs. It is not part of the module lab (which does not need EC2), so treat it as an extra.
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
<TargetFramework>net10.0</TargetFramework>
<Nullable>enable</Nullable>
<ImplicitUsings>enable</ImplicitUsings>
<TreatWarningsAsErrors>true</TreatWarningsAsErrors>
<RootNamespace>Harbourline.AzSurvey</RootNamespace>
</PropertyGroup>
<ItemGroup>
<PackageReference Include="AWSSDK.EC2" Version="4.1.4" />
<!-- Needed at runtime so IAM Identity Center (SSO) profiles resolve. -->
<PackageReference Include="AWSSDK.SSO" Version="4.1.0.2" />
<PackageReference Include="AWSSDK.SSOOIDC" Version="4.1.0.2" />
</ItemGroup>
</Project>using Amazon;
using Amazon.EC2;
using Amazon.EC2.Model;
using Amazon.Runtime.CredentialManagement;
var profileName = Environment.GetEnvironmentVariable("AWS_PROFILE") ?? "learn-sandbox";
var regionName = args.Length > 0 ? args[0] : "eu-west-2";
var chain = new CredentialProfileStoreChain();
if (!chain.TryGetAWSCredentials(profileName, out var credentials))
{
Console.Error.WriteLine($"Profile '{profileName}' not found. Run: aws configure sso");
return 2;
}
using var ec2 = new AmazonEC2Client(credentials, RegionEndpoint.GetBySystemName(regionName));
var response = await ec2.DescribeAvailabilityZonesAsync(new DescribeAvailabilityZonesRequest
{
Filters = [new Filter { Name = "zone-type", Values = ["availability-zone"] }],
});
Console.WriteLine($"Availability Zones visible to '{profileName}' in {regionName}:");
foreach (var az in response.AvailabilityZones ?? [])
{
Console.WriteLine($" {az.ZoneName,-14} {az.ZoneId,-10} {az.State}");
}
return 0;Run it with dotnet run -- eu-west-2. Expected output (your names may map to different IDs):
Availability Zones visible to 'learn-sandbox' in eu-west-2:
eu-west-2a euw2-az2 available
eu-west-2b euw2-az3 available
eu-west-2c euw2-az1 availableCode explained line by line
AWSSDK.EC2โ the EC2 client lives in its own NuGet package; SDK v4 packages shareAWSSDK.Core.AWSSDK.SSOandAWSSDK.SSOOIDCโ not referenced in code, but the SDK loads them at runtime to turn an SSO profile into temporary credentials. Leaving them out compiles fine and fails at runtime (the SDK lesson explains why).Environment.GetEnvironmentVariable("AWS_PROFILE") ?? "learn-sandbox"โ the same profile precedence as the lab: environment first, then the course default.new CredentialProfileStoreChain()โ reads%USERPROFILE%\.aws\configandcredentials, like the CLI does.TryGetAWSCredentials(...)โ builds a credentials object from the profile; for SSO profiles it uses the cached token fromaws sso login.RegionEndpoint.GetBySystemName(regionName)โ convertseu-west-2into the SDK's Region object. Every regional client is bound to one Region; to survey two Regions you create two clients.Filters = [new Filter { ... }]โ a C# collection expression; the filter mirrors the CLI'sName=zone-type,Values=availability-zone.response.AvailabilityZones ?? []โ in SDK v4, collections in responses can benullwhen empty, so guard them.az.ZoneNamevsaz.ZoneIdโ the account-specific label versus the stable physical identifier.- The call needs the IAM permission
ec2:DescribeAvailabilityZones; yourSandboxAdminrole has it.
Security considerations
- Region choice is a security control. Enforce it with an SCP (Module 0) rather than trusting people to pick the right dropdown. Keep the set of enabled opt-in Regions equal to what you actually use.
- Global services need extra care. IAM, Organizations and CloudFront configuration affect every Region. Protect them with least privilege and log them (CloudTrail records global-service events in us-east-1; Module 5).
- Edge caching and personal data. Never cache responses containing customer data at the edge unless the cache key, headers and expiry are designed for it. Statement downloads should use short-lived signed URLs, not public caching.
- Cross-Region copies are data transfers. S3 replication, DynamoDB global tables or backup copy jobs move personal data to another Region; each needs a lawful basis and a data-protection review.
โ Learning shortcut โ not for production
This course defaults to us-east-1 because it is the cheapest common Region, has every service first, and hosts the global control planes, which keeps labs simple. A real UK bank would put UK customer data in eu-west-2 (London) and EU subsidiary data in an EU Region, enforce that with SCPs and resource-level checks, and have legal and the data-protection officer approve any processing elsewhere. Lab data here is fictional, which is the only reason us-east-1 is acceptable.
โ Learning shortcut โ not for production
Using AdministratorAccess to run read-only survey commands is more power than needed. In production, a read-only role (for example one with the AWS managed ReadOnlyAccess policy, or narrower) would run inventory scripts.
Banking use case
Harbourline Bank is launching a Statements Service: customers download monthly PDF statements, and the Data Platform keeps seven years of statement data. The architecture team writes this ADR (all details fictional):
| ADR field | Decision |
|---|---|
| Context | UK retail customers (GBP) are served from the UK bank; EU customers (EUR) belong to the EU subsidiary. Data-protection reference: UK GDPR and EU GDPR. |
| Decision | UK customer data in eu-west-2; EU subsidiary data in eu-west-1. Each Region runs the service across three AZs. CloudFront in front of both for static assets only; statement PDFs are fetched with signed URLs and never cached at the edge. |
| Latency | Customers are in the UK and the EU; both Regions are close. No Local Zone needed. |
| Service availability | All required services (S3, Lambda, DynamoDB, KMS, API Gateway) confirmed in both Regions via SSM public parameters and service docs on the decision date. |
| Cost | S3 storage roughly 4% higher in London than in us-east-1; accepted. |
| Resilience | AZ failure: tolerated automatically. Region failure: recovery plan uses encrypted backups in a second Region chosen by risk and legal (Module 12). |
| Sovereignty | AWS European Sovereign Cloud evaluated for the EU subsidiary; not chosen now because the subsidiary's regulator has not required it โ revisit annually. |
| Review required | Data-protection officer, operational-resilience lead and legal must sign off. This ADR does not by itself make the design compliant with anything. |
Note what the ADR does not claim: it does not say the design "is GDPR compliant". Region choice is one input; lawful basis, retention, access control, contracts and sub-processor review are others.
Common mistakes
- Creating resources in the wrong Region โ the console remembers the last Region you used โ always glance at the Region selector before clicking Create, and keep your SCP tight so mistakes fail fast.
- Assuming
us-east-1ais the same place for everyone โ AZ names are mapped per account in older Regions โ use AZ IDs when two accounts or vendors coordinate. - Building "multi-AZ" with one instance โ people confuse a regional service with a zonal resource โ for zonal resources, run at least two copies in two AZs behind a load balancer.
- Assuming every service and feature exists in every Region โ new features often launch in us-east-1 first โ check availability before committing to a Region.
- Treating data residency as solved by Region choice alone โ logs, backups, support cases and analytics exports can move data โ map every data flow.
- Forgetting cross-Region and cross-AZ data-transfer charges โ transfer between AZs and between Regions is billed per GB โ keep chatty traffic inside one AZ where safe, and price replication explicitly.
- Picking the cheapest Region for customer data โ saves cents per GB โ can create legal exposure worth far more.
Troubleshooting
| Symptom | Likely cause | Fix |
|---|---|---|
AccessDenied ... with an explicit deny in a service control policy | The Region guardrail SCP blocks that Region | Use us-east-1 or eu-west-2, or (in production) change the SCP through review |
OptInRequired or "The security token included in the request is invalid" only in one Region | The Region is opt-in and not enabled for this account | Enable it in Global View (and allow it in the SCP), or choose another Region |
| Your bucket, function or table "disappeared" | The console is pointed at a different Region | Switch Region, or use Global View / Tag Editor to search all Regions |
describe-availability-zones shows fewer AZs than the documentation | Constrained AZs are hidden from newer accounts | Expected; design for the AZs you can see |
get-parameters-by-path returns nothing | Path typo, or querying a Region that does not support global-infrastructure parameters | Check the path, add --recursive if needed, and query from us-east-1 |
| CloudFront or ACM certificate errors when the bank uses only eu-west-2 | CloudFront certificates must be in us-east-1 | Allow the specific ACM actions in us-east-1, or accept the global-service exception in your ADR |
Cost and cleanup
Cost: $0. Every command in this lesson only reads metadata. describe-regions, describe-availability-zones, Account API calls and reading AWS public SSM parameters have no charge. The optional C# app creates nothing.
Cleanup: nothing to delete. If you enabled an opt-in Region while exploring, disable it again so the guardrail and the account agree:
# Only needed if you enabled a Region during this lesson. Run from an identity with account:DisableRegion.
aws account disable-region --region-name eu-south-1 --profile learn-sandbox
aws account get-region-opt-status --region-name eu-south-1 --profile learn-sandboxExercises
- ๐ข Guided: Run CLI commands 1โ5. Fill in a table of ZoneName โ ZoneId for us-east-1 and eu-west-2, and copy the exact text of the SCP denial from command 5.
- ๐ก Challenge: Using only the SSM public parameters, produce a list of Regions where both AWS Lambda and Amazon DynamoDB are available, then check which of those are enabled by default. (Hint: two queries and a comparison in PowerShell with
Compare-Object, or in C#.) - ๐ด Live problem: A product owner writes: "Our partner in Ireland says their AZ
eu-west-1ais where they run the card-tokenisation service, so let's put our app ineu-west-1atoo for lowest latency. Also, can we move statements to us-east-1 to save money?" Write the reply you would send.
Assignment
Add docs/region-decision.md to your learning-journal repository: an ADR for a fictional Harbourline workload of your choice (for example the Payments Service). Include the decision table from the banking use case, the CLI output that proves service availability, a price comparison for one service across two Regions with the date you checked, and a "who must approve" section.
Solutions
You get six AZs in us-east-1 and three (sometimes four) in eu-west-2. The IDs (use1-az1โฆ, euw2-az1โฆ) are stable; the letter suffixes may not line up with the ID numbers. Command 5 fails with a message containing "explicit deny in a service control policy". Why: the SCP evaluates aws:RequestedRegion on every call; describe-availability-zones sent to eu-central-1 carries that Region and is denied, while describe-regions is exempt in the guardrail's NotAction list.
$lambda = aws ssm get-parameters-by-path --region us-east-1 --path /aws/service/global-infrastructure/services/lambda/regions --query "Parameters[].Value" --output text --profile learn-sandbox
$ddb = aws ssm get-parameters-by-path --region us-east-1 --path /aws/service/global-infrastructure/services/dynamodb/regions --query "Parameters[].Value" --output text --profile learn-sandbox
$both = ($lambda -split '\s+') | Where-Object { ($ddb -split '\s+') -contains $_ } | Sort-Object
$default = aws account list-regions --region-opt-status-contains ENABLED_BY_DEFAULT --query "Regions[].RegionName" --output text --profile learn-sandbox
$both | Where-Object { ($default -split '\s+') -contains $_ }The SSM results include opt-in Regions; the second filter keeps only Regions your account has by default. Why: service availability and account availability are separate questions โ a Region can offer a service your account cannot reach yet.
Key points of a good reply:
- AZ names are not shared. The partner's
eu-west-1amay be a different physical AZ from yours. Ask for their AZ ID (for exampleeuw1-az2) and map it withdescribe-availability-zonesin your account. - Don't pin to one AZ anyway. Latency between AZs in a Region is low (single-digit milliseconds); deploying in a single AZ to save a millisecond creates a single point of failure. Run across at least two AZs and, if latency truly matters, prefer the partner's AZ ID for the primary.
- Moving statements to us-east-1 saves about $0.001 per GB-month on S3 storage (about 4%) but moves UK and EU customer personal data to the US. That needs a lawful transfer mechanism, a data-protection impact assessment and DPO/legal approval โ and it contradicts the Region ADR. Recommend lifecycle rules to cheaper S3 storage classes in the same Region instead (Module 6).
Why: the ticket mixes a misunderstanding (AZ names), a resilience risk (single AZ) and a compliance risk (data residency). Good engineers separate them.
Knowledge check
Answer: Three. AWS designs every Region with at least three AZs so that you can survive the loss of one and still have two.
Answer: No. In older Regions, AZ names were mapped to physical AZs independently for each account (accounts created from November 2025 get a consistent mapping). The AZ ID is the stable identifier to compare.
Answer: IAM role โ global. S3 bucket โ regional (with a partition-wide unique name). EBS volume โ zonal. Lambda function โ regional.
Answer: No. It is a separate partition (aws-eusc) with its own accounts, IAM and billing. A standard-partition account cannot list or access its Regions.
Answer: They serve CloudFront (caching, TLS termination, routing to the origin) and some other edge services. You do not deploy your API servers there; the API runs in a Region, and CloudFront fronts it.
Interview questions
- "How do you choose an AWS Region for a new banking workload?" โ Start with data-residency and regulatory requirements (they are constraints, not preferences), then confirm service and feature availability, then latency to users and partners, then price. Record it in an ADR and get legal/DPO sign-off.
- "What's the difference between high availability and disaster recovery in AWS terms?" โ HA usually means surviving an AZ failure within a Region automatically (multi-AZ). DR means recovering from a Region-level event, using a second Region with defined RTO/RPO; it is never automatic unless you build it.
- "Why might two teams disagree about which AZ a service is in?" โ AZ name mapping per account; use AZ IDs.
- "Name a global AWS service and explain a risk that creates." โ IAM: a change applies everywhere instantly, and its control plane lives in us-east-1, so SCPs must exempt it when restricting Regions.
Certification practice
A company wants to reduce latency for users downloading static images worldwide, without deploying servers in new Regions. Which AWS infrastructure component should it use?
A. Availability Zones B. Edge locations through Amazon CloudFront C. AWS Outposts D. Additional VPCs
Answer: B. CloudFront caches content at edge locations near users. A: AZs are inside Regions, not near every user. C: Outposts put AWS hardware in your own data centre โ not a global delivery network. D: VPCs are regional networks and do not reduce global latency.
Which design gives an application high availability within a single AWS Region?
A. Run two EC2 instances in the same Availability Zone B. Run instances in at least two Availability Zones behind a load balancer C. Run one large instance with more CPU D. Use an edge location as a backup server
Answer: B. Spreading across AZs survives the loss of a data-centre group. A shares the AZ's failure domain. C scales up but remains a single point of failure. D: edge locations do not host your servers.
A UK bank must keep customer data in the United Kingdom. Developers in its AWS Organization sometimes create resources in other Regions by mistake. Which combination best enforces the requirement? (Choose TWO.)
A. Attach an SCP to the workload OUs that denies actions when aws:RequestedRegion is not eu-west-2, exempting required global services
B. Use CloudFront to cache all customer data in UK edge locations
C. Keep opt-in Regions disabled for the workload accounts
D. Ask developers to check the Region selector before creating resources
E. Store data in us-east-1 and encrypt it with a KMS key in eu-west-2
Answer: A and C. A is a preventive guardrail across all accounts. C removes opt-in Regions entirely. B moves data into caches and is not a residency control. D is guidance, not enforcement. E: a KMS key in another Region cannot directly encrypt S3 objects in us-east-1, and the data would still be in the US.
A company shares subnets from a networking account with a workload account. The workload team says its servers in eu-west-1a see higher latency to a database than expected, even though the database is "also in eu-west-1a". What is the most likely explanation?
A. The database is in a Local Zone
B. AZ names are mapped independently per account, so the two eu-west-1a names refer to different physical AZs
C. eu-west-1 has only one AZ
D. Cross-account traffic always crosses Regions
Answer: B. In older Regions each account has its own name-to-AZ mapping; compare AZ IDs instead. A is possible in theory but would have a different zone name. C is false; every Region has at least three AZs. D is false; cross-account traffic within a Region stays in that Region.
Completion checklist
- I can define Region, AZ, edge location, Local Zone, Wavelength Zone, Outpost and partition in one sentence each
- I listed the AZ names and IDs for us-east-1 and eu-west-2 in my sandbox
- I saw the SCP block a non-approved Region and can explain the error
- I checked service availability with SSM public parameters
- I wrote a Region ADR for a Harbourline workload
Key takeaways
- Partitions contain Regions; Regions contain at least three AZs; AZs contain data centres. Edge locations, Local Zones, Wavelength Zones and Outposts extend the network outward.
- The Region is your data-residency boundary; the AZ is your failure boundary. Multi-AZ is how you get high availability; it is not automatic for zonal resources.
- Use AZ IDs, not AZ names, when coordinating across accounts.
- Choose a Region on residency and regulation first, then services, latency and price โ and document the decision.
- Enforce Region choices with SCPs and opt-in settings, not with good intentions.
Further reading
- AWS global infrastructure
- AWS Regions (list, opt-in status, CLI examples)
- AWS Availability Zones and AZ IDs
- Enable or disable AWS Regions in your account
- Public parameters for Regions, services and AZs (Systems Manager)
- AWS Local Zones locations
- Amazon CloudFront features (edge network)
- Opening the AWS European Sovereign Cloud (AWS News Blog)
- AWS data privacy FAQ
Independent learning resource โ not affiliated with, endorsed or sponsored by Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates. All banking data in examples is fictional. Sample architectures are for learning and are not automatically compliant with GDPR, PCI DSS or any banking regulation.
Public, private and hybrid cloud
Explain the cloud deployment models, why banks run hybrid for years, and which AWS building blocks connect an on-premises data centre to AWS for networking, identity, compute and data residency.
AWS accounts and Organizations
Use AWS accounts as security, billing and quota boundaries, organise them with AWS Organizations and OUs, and choose the right organisation policy for a bank's guardrails.