What is fault domain?
In the physical world, if a water pipe bursts in your kitchen, your bathroom might still work. But if the main water line for the entire street breaks, every house on the block is in trouble.
In cloud computing, a Fault Domain is a set of hardware components—servers, switches, and power supplies—that share a single point of failure. Understanding fault domains is the key to building an application that stays online even when hardware breaks.
A fault domain is a "blast radius." If a specific piece of equipment fails, everything within that fault domain is affected.
Cloud providers organize their data centers into these domains to ensure that a single "oops" doesn't take down an entire region. Usually, a fault domain consists of:
A Rack of Servers: Multiple physical computers.
A Power Source: The PDU (Power Distribution Unit) feeding those servers.
A Network Switch: The hardware that connects those servers to the internet.
The Core Rule: If you put two copies of your website on two different servers in the same rack, and that rack’s power supply fails, your website is 100% offline. To be safe, you must spread your servers across different fault domains.
Cloud providers often talk about these two concepts together, especially in platforms like Microsoft Azure:
| Feature | Fault Domain (FD) | Update Domain (UD) |
| Nature | Unplanned (Hardware Failure) | Planned (Software Maintenance) |
| Cause | Power outage, broken switch, fire | Security patches, OS updates |
| Goal | Protect against "Acts of God/Hardware" | Protect against "Acts of the Provider" |
Update Domains ensure that when the cloud provider needs to reboot servers to install a security patch, they don't reboot all your servers at the exact same time.
When you deploy an application, you use Availability Sets or Anti-Affinity Rules to tell the cloud provider how to distribute your "workers."
The Request: "I want to run 3 web servers."
The Logic: The cloud orchestrator looks at its map and places Server A in Fault Domain 1, Server B in Fault Domain 2, and Server C in Fault Domain 3.
The Result: If a rogue forklift in the data center crashes into the rack for Fault Domain 1, your website stays online because Servers B and C are in completely different physical locations with their own power and networking.
Fault domains are the "micro" level of cloud resilience. Most cloud providers follow this hierarchy:
Fault Domain: A single rack/power supply.
Availability Zone (AZ): A whole building (data center). Multiple fault domains live inside an AZ.
Region: A geographic area (like "US-East"). Multiple AZs live inside a Region.
SLA Guarantees: Most cloud providers only guarantee "99.99% uptime" (the Service Level Agreement) if you distribute your instances across multiple fault domains. If you put everything in one domain, you are on your own.
Cost vs. Risk: Spreading across fault domains is usually free—it’s just a configuration setting. Failing to do it is like owning a car and refusing to wear a seatbelt because it "takes too much effort" to click it in.
A fault domain is a boundary of failure. By acknowledging that hardware will eventually break, you can design your system to be "Fault Tolerant." You don't try to prevent the hardware from failing; you simply make sure your application is spread out enough that no single failure can take the whole thing down.