Hacker News Reader

Comment by nemothekid | Hacker News Reader

I've been naively setting up our distributed databases in separate AZs for a couple years now, paying, sometimes, thousands of dollars per month in data replication bandwidth egress fees. As far as I can remember I've never never seen an AZ go down, and the only region that has gone down has been us-east-1.

View on HN

There was an AZ outage in Oregon a couple months back. You should definitely go multi AZ without hesitation for production workloads for systems that should be highly available. You can easily lose a system permanently in a single AZ setup if it’s not ephemeral.

AZs definitely go down. It's usually due to a physical reason like fire or power issues.

> I've never never seen an AZ go down, and the only region that has gone down has been us-east-1.

Doesn't the region going down mean that _all_ its AZs have gone down? Or is my mental model of this incorrect?

No. See https://aws.amazon.com/about-aws/global-infrastructure/regio...

A region is a networking paradigm. An AZ is a group of 2-6 data centers in the same city more or less.

If a region goes down or is otherwise impacted, its AZs are unavailable or similar.

If an AZ goes down, your VMs in said centers are disrupted in the most direct sense.

It's the difference between loss of service and actual data loss.

Is that separate AZs within the same region, or AZs across regions? I didn't think there were any bandwidth fees between AZs in the same region.

It's $0.01/GB for cross-AZ transfer within a region.

In reality it's more like $0.02/GB. You pay $0.01 on sending and $0.01 on receiving. I have no idea why ingress isn't free.

Plus the support percentage, don’t forguet.

That is incorrect. Cross az fees are steep.

Reply on news.ycombinator.com