Despite the “cloud” metaphor explained in What Is Cloud Computing?, every workload you run with a cloud provider executes on physical servers in a specific physical location. Understanding how providers structure those locations — into regions and availability zones — is basic to reasoning about latency, resilience, and cost, not an advanced topic.

Regions: where your data actually lives

A region is a distinct geographic area where a cloud provider operates data center infrastructure — for example, a specific area of the eastern United States, or a specific part of Western Europe. Providers maintain many regions worldwide, each operating largely independently of the others.

Region choice matters for a few concrete reasons:

  • Latency — network distance to your users has a physical floor. A workload serving European users from a North American region will have meaningfully higher latency than one running in a European region.
  • Data residency and compliance — some regulatory regimes require certain categories of data to stay within specific geographic or legal jurisdictions, which constrains which regions are viable.
  • Pricing — compute, storage, and network pricing can differ meaningfully by region for the same underlying service.
  • Service availability — not every service is available in every region, particularly newer ones.

Availability zones: resilience within a region

Within a region, providers typically operate multiple availability zones (AZs) — physically separate data centers with independent power, cooling, and networking, but connected by low-latency links fast enough to run tightly coupled systems across them. AWS describes this structure directly in its overview of regions and availability zones: each AWS region is designed to be completely isolated from other regions, and each contains multiple AZs designed so a failure in one AZ is unlikely to affect the others.

This is the mechanism behind a common piece of cloud architecture advice: distribute redundant instances of a service across multiple AZs within a region, not just multiple servers within a single data center. A single AZ can still fail — power outages and network incidents happen — and if your entire deployment sits in one AZ, that failure takes your service down with it. Spreading replicas across AZs, and configuring load balancing and failover to route around a failed AZ automatically, is one of the most basic resilience patterns in cloud architecture, and it depends entirely on understanding this geographic structure.

What this means for architecture decisions

Region and AZ structure directly shapes decisions covered elsewhere on this site. It’s a major input into the cost decisions discussed in Why Cloud Bills Get Out of Control — cross-region and, in some cases, cross-AZ data transfer is often billed, and region choice affects unit pricing. It also underpins the reliability targets discussed in Understanding SLIs, SLOs, and Error Budgets: an availability target that assumes single-AZ deployment is architecturally capped by that AZ’s own failure rate, no matter how well the application code is written.

Multi-region: a bigger step than multi-AZ

Running across multiple AZs within a region protects against a data-center-level failure. Running across multiple regions protects against a regional-level event, and against a specific region becoming unavailable, but it’s a substantially bigger architectural undertaking — typically requiring cross-region data replication, and careful thought about consistency and failover, rather than something that can be turned on with a configuration flag. Most services don’t need multi-region resilience; the cost and complexity are only worth it when a regional outage is a risk your specific availability requirements can’t tolerate.

Key takeaway

Cloud regions and availability zones are the physical foundation everything else in cloud architecture is built on: region choice sets your latency floor, data residency options, and pricing, while distributing workloads across availability zones within a region is the standard, comparatively low-cost way to survive a single data center failure. Multi-region resilience solves a different, larger problem and should be reached for deliberately, not by default.

This article explains general cloud infrastructure concepts. Specific region names, AZ counts, and service availability vary by provider and change over time — verify current details against provider documentation before making architecture decisions. See our disclaimer.