Is it time to diversify your AWS cloud?
If you missed the news, last week AWS had a major outage in it’s east coast datacenters.
What does this news mean for your business?
Join 35,000 others and follow Sean Hull on twitter @hullsean.
If you’re like a lot of customers – and I mean a *lot* of them, you probably have resources deployed in us-east-1. And if you haven’t done it right you probably had an outage last week.
1. How does AWS organize their cloud?
Amazon Web Services, not to be confused with the bookseller side of the business, provides on-demand compute power to customers. Want to run an application on the internet? Chances are you’re looking to AWS to host those servers.
They’ve designed a brilliant global network of datacenters, which are organized into regions, and within those regions individual buildings or datacenters themselves. Inside those buildings are rows and rows and rows of interchangeable servers, that you configure without ever visiting the building.
Those regions are named based on where they are in the world. us-east-1 (north virginia) us-east-2 (ohio), us-west-1 (Oregon) and so on. There are many many regions globally. Within those you have individual datacenters or availability zones. For example in northern virginia you have us-east-1a, us-east-1b, us-east-1c and so on.
Read: Is AWS too complex?
2. What happened?
AWS had an outage of the US-EAST region. Customers deployed in that region were affected. Applications went offline, maybe data was lost.
As it turns out North Virginia is the oldest region. That’s where it all began for AWS.
Want a bit more history, reach Why Amazon’s data centers are hidden in spy country.
3. Which customers were affected?
There are two ways you could have been affected by this outage.
1. You deployed in the default region and figured that was a safe bet
2. You deployed *only* in one region
These are two no-nos of cloud deployments, but a lot of customers still do it. Perhaps out of ignorance, or out of pressure to get it done.
4. How can I mitigate these types of outages?
The right way to deploy is to use devops automation for starters. This means everything is in code. Want to move your full stack to another AZ or region? Easy just change one variable, and redeploy. A short while later all your resources will be moved over, and you’ll be back up and running. That’s powerful.
You can also use various types of DNS tricks. Cross-zone load balancing, is very powerful, and allows you to deploy some assets in different zones, and therefore avoid outages affecting individual zones.