What are cloud regions and availability zones?


If you use a public cloud, you’ve used availability zones (AZ), but what are they really? How do they fit into regions anyway? Let’s go over the basics.

First, a region is not a data center. No, I know what you’ve been told, but it’s not. A region must have at least one data center within it, but they’re often made up of multiple data centers. For example, Amazon Web Services (AWS)‘s US East region is made up of six data centers in northern Virginia.

Each region is designed to be completely isolated from the other regions. If a Richter scale 8 earthquake smacked AWS’s US West region, the other regions would keep running.

Not all regions are created equally. AWS US East, for example, tends to have more downtime than its West coast counterparts. That’s because US East, the oldest AWS region, tends to also have the oldest infrastructure.

If you want to really ensure your applications will keep running this side of armageddon, you’ll want to co-locate them in multiple regions.

There are also services, such as Amazon Route 53, which enable sysadmins or DevOps programs to redirect IP traffic from a failing or overloaded region into an operational one. AWS’s Elastic Load Balancer can also automatically distribute traffic to applications, run with Auto Scaling, to ensure a particular workload has the computing power it needs to meet demand.

Each region, like the name indicates, is a separate geographic area. Within each region there are multiple, isolated locations: Availability Zones. Each AZ has at least one data center. AZs have independent power sources, networking, and cooling resources. Typically, especially if you’re new to the public cloud, your presence will be on a single AZ.

A single AZ can run on multiple data centers, but no two zones share a single data center. That’s because, as Amazon puts it, “customers who care about the availability and performance of their applications want to deploy these applications across multiple Availability Zones in the same region for fault tolerance and low latency.”

With an AZ, its data centers are hooked up to each other over redundant low-latency, high-speed fibre network links. Typically, these run at 40Gbps using InfiniBand or Ethernet interconnects. All the zones in a region link with each other over redundant private network links. The internet connects you with the cloud, but clouds themselves are bound by private connections. With the endless demand for speed, especially for data-replication between AZs, the major public cloud providers are moving to software-defined networks (SDN) 100Gbps network fabric.

AWS is not the only cloud provider to use the region/AZ framework for its cloud. For instance, both Google Compute Engine (GCE) and IBM Cloud use this approach.

For the longest time, Microsoft Azure did not use regions and AZs. Instead, Microsoft just used regions. In the older Microsoft scheme a region might have only a single data center. In 2017, Azure decided to adopt the AZ method as well. At the end of March, Microsoft started rolling out production AZs.

Looking ahead, it’s clear regions and AZs are the model for all major public clouds. It’s possible that with the advent of Kubernetes, which enables hybrid clouds, we’ll see other cloud deployment models appear, but I’m not betting on it.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: