Datacenter Power Outage? Stay online!

Posted July 5, 2012 at 3:49 pm by Justin van Heerde in Cloud Computing, General, Infrastructure.

Last Friday, Amazon’s Northern Virginia datacenter (US-EAST) had a power failure because of electrical storms in the area. This is the second time the same datacenter gets hit by a major power failure, and several high profile sites and services were impacted: Netflix, Reddit, Pinterest, Instagram.

So, what does this mean? Does it mean we can’t trust Amazon’s EC2 service, still the most advanced public IaaS offering in the market today, to host our business critical services? Should you just avoid using ‘the cloud’ altogether? No and no.  It all comes down to the way you design and manage your business critical cloud IT infrastructure.

Ask yourself this: when you decide to outsource the hosting of your services, would you want to depend on a single datacenters’ power facility? Of course not, you would set up everything redundantly across different datacenters. Also, are other datacenters immune to natural disasters? Again, no. The storm that took out (parts of) Amazon’s data center also took out power for 2 million people in the DC area.

So what can we do about events like this taking out your business for hours at a time? It’s all in application design. Amazon offers different availability zones per region. By setting up services across availability zones and regions you can make sure your services stay online. Better yet, you could set up your infrastructure across several different cloud providers, also known as cloud-over-cloud. That way you won’t be exclusively dependent on Amazon. Just ask Jitscale about this cloud-over-cloud service and we’ll gladly help you with designing your ultra-resilient cloud-based infrastructure.