Your Cloud Broke. Now What?
As I'm writing this, the Internet is finally recovering from the massive outage experienced by Amazon.com's cloud computing infrastructure. Wow, you just don't even realize who's hosting with them until something goes wrong: HootSuite, NetFlix, Reddit, FourSquare, and many more major names all experienced outages.
It's okay -- computers break. I'm sure the folks at Amazon are looking hard at why and coming up with ways to prevent that kind of failure again. Great -- let's move forward.
Actually, let's look back for just a second. Not at the failure or the fallout, but what this means to a business who has chosen to outsource to a cloud computing provider. I know that Amazon provides a solid Service Level Agreement (SLA) to their customers, and that the customers affected by this outage will doubtless be receiving a service credit. Of course, since you "pay as you go" with this kind of cloud computing model, you could argue that these customers aren't owed very much, because they weren't consuming any resources.
Therein lies a risk for cloud computing, and it's one that pundits have identified in the past: That SLA will never cover your lost business. I don't know how much business NetFlix lost, or HootSuite (which has paying customers to contend with, now), or any of the others -- but I know their business impact won't even begin to be covered by whatever Amazon gives them under their SLA.
This has been a major argument for avoiding cloud computing services. I think, however, that it's a bad argument. Consider the alternative: You host all of your services in your own data center. It goes down, like Amazon did, and your redundancies don't work either, which is what happened to Amazon. Who's going to compensate you for your business loss?
Nobody. Certainly not Dell, IBM, HP or whoever sold you your servers. Not the power company, if that was the source of the problem. Not your HVAC contractor, if that's what went wrong. Nobody. The only difference is that instead of having the possibility of controlling the situation and repairing it faster, with cloud hosting you just sit and wring your hands and cuss a lot. But notice that I wrote "the possibility." Just because your services are in your own datacenter doesn't mean you will definitely be able to impact the problem in any way. What if your datacenter flooded? Got hit by a meteor? All of your HVAC units burned out? There are plenty of things that go on in your own data center that are beyond your control, and when those things go wrong you're still tearing hair, cussing, etc.
Yes, you can build redundancies for almost anything -- up to and including a complete, redundant datacenter. Cloud hosting providers can do that, too, and it's often more economical for them. So I'd argue that you simply need to select a hosting provider that has built a more redundant system than you could afford to do yourself. The resources-on-demand model of cloud computing is really, really compelling for a number of businesses. There are still valid reasons not to outsource to a cloud computing provider, such as data security concerns, but I don't think the "well, the SLA won't pay us for our business loss" is a valid reason. Do your due diligence and make sure your provider is set up the way you like, and then hope for the best. That's all you do in your own data center.
Posted by Don Jones on 04/29/2011 at 1:14 PM