By now, you’ve no doubt heard about yesterday’s Amazon Web Services outage (tracked on Twitter with the inauspicious hashtag #AWSpocalypse), during which a power outage disrupted cloud service delivery from its Northern Virginia data center, leaving many customers (as well as Pinterest, Salesforce.com’s Heroku and other high-traffic web services) out in the cold. In lieu of an official post-mortem from AWS, let’s step back and look at what we’ve learned from all this.
Analysts, vendors, and armchair CIOs are coming forward proclaiming this as the final nail in the coffin for public cloud. Piston Cloud, provider of private cloud infrastructure services built on the OpenStack platform, generated plenty of buzz around a blog published today entitled “AWS Outage Proves It’s Better to Own Than Rent.”
You can probably guess the thrust of Piston co-founder Gretchen Curtis’ argument from the name alone, but essentially the company is putting forward the idea that public cloud services simply aren’t reliable or suitable enough for mission-critical workloads, and this outage just underlines the benefits and cost savings of a private cloud that you own.
The response to that one blog entry has been tremendous, with a lot of debate around the public vs. private cloud cost curve, and when it makes better economic sense to move from public cloud services back behind your firewall. Fans of the public cloud shot back that private data centers are just as prone to failure, and with the added expense and time sink of infrastructure design and monitoring.
Rackspace Hosting, which competes with Amazon Web Services on the infrastructure-as-a-service front, kept it a little classier with a Twitter post that read “We wish our friends at AWS and all of their customers the best. Everyone suffers down-time and never at a good time.”
It’s good that Rackspace took it upon itself to raise the tone, because as enStratus CTO George Reese put it on his Twitter account, this kind of schadenfreude just hurts the public cloud services market as a whole, especially since most competitors are far from blameless.
The kind of CIO who has severe misgivings about the cloud isn’t going to be likely to take the time to distinguish between public vs. private – they’re just going to dismiss the entire concept out of hand. Of course, Reese also tweeted directly at Piston’s Curtis, calling her blog post a lie, so make of that what you will.
No, if there’s a lesson here, it’s not that Amazon Web Services is inherently unreliable, or that private clouds are superior to public clouds. It’s the idea that the market needs to work even harder at designed-to-fail cloud infrastructure with failover and redundancy, and that developers and customers need to work towards the kind of multi-cloud future that PaaS evangelists from VMware Cloud Foundry, AppFog and others have been telling us about for some time.
If you’re deploying a mission-critical application, it’s increasingly vital that you either ensure resiliency across availability zones (in the case of Amazon Web Services) or across multiple cloud infrastructure vendors. It’s just a matter of time before any cloud provider experiences some kind of service disruption. But the odds of all of them suffering an outage independently are very, very slim. Geographic dispersion saves lives, or at least data.