Availability Zone
An availability zone (AZ) is one or more physically separate data centers within a cloud region, each with independent power, cooling, and networking, designed so that a failure in one zone does not take down the others.
Definition
An availability zone (AZ) is one or more physically separate data centers within a cloud region, each with independent power, cooling, and networking, designed so that a failure in one zone does not take down the others.
Overview
Availability zones are the fault-isolation layer that sits beneath cloud regions. While a region represents a broad geography, it's typically composed of two or more availability zones — physically distinct facilities, often several kilometers apart and on separate power grids and network paths, but connected by low-latency, high-bandwidth private links so they can function together as a cohesive regional platform. The purpose of this separation is to contain the blast radius of infrastructure failures: a power outage, cooling failure, or hardware fault in one AZ should not cascade to others in the same region. Cloud architects use availability zones as the primary building block for high availability within a single region: deploying redundant instances of an application across two or three AZs, fronted by a load balancer, means the loss of one entire zone still leaves the application serving traffic from the remaining zones. Managed services like RDS Multi-AZ deployments, Kubernetes node pools spread across zones, and auto-scaling groups with multi-AZ configuration all build on this pattern. This is distinct from multi-region architecture — spreading across AZs protects against data-center-level failures within one geography, while spreading across regions protects against broader, geography-level disruptions, at the cost of higher latency and complexity for cross-region replication. Because AZs within a region are connected by low-latency private networking, applications can typically treat multi-AZ deployment as a relatively low-cost way to gain resilience — much cheaper in complexity and latency overhead than multi-region — which is why 'deploy across at least two/three AZs' is a standard baseline recommendation from every major cloud provider for anything beyond a toy workload.
Key Concepts
- Physically isolated data center(s) within a single cloud region
- Independent power, cooling, and networking per zone to contain failure blast radius
- Connected to other AZs in the same region via low-latency, high-bandwidth private links
- Primary building block for single-region high-availability architectures
- Supported natively by managed services (e.g. Multi-AZ RDS, zone-aware Kubernetes scheduling)
- Lower latency/complexity overhead than multi-region redundancy
- Load balancers can distribute traffic across healthy AZs automatically
- Typically 2-6 AZs available per region depending on the provider and region