AWS
Intermediate 14 min read AWS

AWS Global Infrastructure

Regions, AZs and edges — where things actually run

AWS regions, AZs, local zones and edge locations form a geography of isolation and acceleration. Knowing which layer hosts what matters for latency, residency and DR.

AWS regionsavailability zoneslocal zonesoutpostsedge location
Prerequisites: Regions and Availability Zones
Before we start

What you will learn

  • Explain AWS's region/AZ/edge hierarchy
  • Match deployment layers to use cases
  • Apply region placement for residency and latency requirements
Introduction

What is this?

AWS geography is layered. Regions contain at least 2 physically isolated AZs; AZs are data centers with independent power/cooling/network connected by low-latency links. Local Zones extend regions closer to users; Wavelength zones embed AWS at 5G networks; Outposts put AWS infrastructure on your premises; CloudFront edge locations cache content at hundreds of points of presence. Data stays in its region unless you deliberately replicate.

Motivation

Why do you need it?

  • AZ awareness turns redundancy into real availability — no single facility failure takes you down.
  • Regions control latency (physics) and residency (law) simultaneously.
  • Local zones/outposts serve ultra-low-latency or on-prem-required workloads that can't live in a regional data center.
  • Edge locations give your CDN and global accelerator speed without hosting durable state.
Think about it

Mental model

Imagine

A city, its hospitals, and ambulance outposts

Think of

Region, AZs, and edge locations

The city (region) covers residency; independent hospitals (AZs) mean one fire can't take the whole city down; outposts near where people are (edges) shorten response times.

Mechanics

How it works

  • You pick a region per workload; services are regional (buckets, DNS) or global (IAM, CloudFront) depending on service.
  • Within a region, distribute instances across at least 2 AZs so capacity survives one AZ's failure.
  • AZ interconnect uses low-latency private links, which is why synchronous database replicas work across AZs.
  • Local zones/outposts present region-like APIs but place hardware closer or on-prem.
  • Edge locations accelerate via Route 53/CloudFront/Global Accelerator but never hold your source of truth.

AWS geographic layers.

Region
AZ-A
AZ-B
Local Zone / Outposts
Edge locations
Real world

In production

  • A SaaS runs its API in ap-southeast-2 spread over two AZs, with CloudFront caching static assets at edge PoPs worldwide.
  • A telco embeds a 5G app in a Wavelength zone to keep gaming round-trips under 20ms.
  • A factory runs Outposts on-site for machine-data capture, syncing aggregates to the parent region.
Guidelines

Best practices

  • Place state where compliance says it must live; place compute near the state, not near you.
  • Design every regional service to survive one AZ loss and test that in a game day.
  • Use edge layers to accelerate, never to store the authoritative copy.
  • Review the region map annually — new regions and zones appear and change the trade-offs.
Watch out

Common mistakes

  • Putting replicas in the same region 'because it's cheaper' — data residency says otherwise.
  • Confusing 'multi-region' with global redirects in a single-region active design.
  • Using one AZ for instance fleets and hoping for the best.
  • Choosing a region by habit while users are on another continent.
Root cause analysis

Common issues

Error
CloudFront URL resolves but content is stale
Root cause

Edge cached an old version past TTL

Fix

Invalidate the cache or use versioned object keys (cache-busting) for updates.

Error
Replicas all fail at once
Root cause

Instances falsely promised multi-AZ but share one subnet

Fix

Verify each replica's actual AZ/subnet and spread them across at least 2 AZs.

Error
Local zone workload can't reach regional services
Root cause

Local zones need explicit VPC routing to the parent region

Fix

Check route tables and security groups for the parent-region path.

When things break

Debugging tips

  • Measure latency from the user's region, not from your office.
  • On an AZ anomaly, shift traffic by adjusting Route 53/LB weights rather than redeploying.
  • Verify replicas are in distinct AZs by listing actual placement, not by trusting a label.
  • Check whether a service is regional or global before troubleshooting cross-region surprises.
Interviews

Interview checkpoint

Test yourself

Quick quiz

0/2
1

Which AWS layer caches content near users but hosts no authoritative data?

2

Data automatically replicates between AWS regions.

Go deeper

Official documentation

AWS — Global infrastructure
docs
AWS — Regional services reference
docs