AWS Global Infrastructure
How Amazon Web Services is laid out physically: a hierarchy of Regions, Availability Zones and edge locations, plus several extensions that push AWS infrastructure closer to users or into a customer’s own data center. Almost every design question about latency, durability or data residency resolves to a choice at one of these levels.
Key points
- Approximate current scale — per AWS’s own Global Infrastructure page (see Sources), approximately 39 Regions, 123+ Availability Zones, 45 Local Zones and 33 Wavelength Zones, plus 750+ CloudFront edge locations (points of presence) and 15 regional edge caches. AWS is continuously building, so treat these as directional, not fixed — check the source page for the current count.
- Region — a physical location containing a cluster of data centers, identified by codes such as us-east-1 (N. Virginia), us-west-2 (Oregon), ap-south-1 (Mumbai), eu-central-1 (Frankfurt). Most services are region-scoped, and a region is chosen for latency, cost, compliance and service availability.
- Availability Zone (AZ) — one or more discrete data centers within a Region with independent power, cooling and networking. Deploying across AZs is the standard way to get high availability; Amazon VPC subnets each live in exactly one AZ.
- Edge location — a global network of sites used to cache and terminate traffic close to end users. Amazon CloudFront delivers content from them, and Amazon S3 Transfer Acceleration and AWS Global Accelerator use them to shorten the path into AWS.
- Regional vs global services — Amazon Route 53, Amazon CloudFront, AWS Identity and Access Management (IAM) and AWS Key Management Service (KMS) behave globally, while Amazon Config, Amazon EC2 Auto Scaling and most compute and database services are configured per region.
- AWS Local Zones — an extension of a Region to a more granular metropolitan location, placing compute and storage nearer to the client for single-digit millisecond latency.
- AWS Outposts — AWS-owned racks installed in the customer’s own data center, running native AWS services on-premises with the same management and control plane. Used where data must physically stay on site or where latency to local systems matters.
- Snow family devices — physical appliances for moving data in and out of the Region when the network is the bottleneck. See AWS Snowball.
- Durability and replication follow the hierarchy — Amazon S3 spreads objects across AZs within a Region, Amazon Aurora keeps six copies across three AZs, and cross-Region replication is always an explicit, separately configured choice. Amazon RDS Multi-AZ synchronously replicates to a standby in a second AZ and fails over to it automatically without changing the endpoint; cross-Region read replicas are a separate, asynchronous mechanism.