The cloud is dramatically different from any IT environment that has gone before. And yet it is remarkable how similar it is to standard infrastructure that we all rely on so heavily.
One of the key similarities, of course, is that like traditional IT, the cloud goes down every once in a while. As Amazon users will tell you, sometimes it goes down quite often. Just weeks after a power failure sent AWS customers scrambling for service elsewhere, the massive East Coast storms took the service out again over the weekend, sparking yet another debate about cloud reliability and the need to provide adequate backup and failover infrastructure should the need arise.
It may come as a surprise to some, but there is still a high degree of wishful thinking when it comes to the efficacy and reliability of the cloud. As Alcatel-Lucent found out in a recent survey, close to two-thirds of top IT executives are looking for guaranteed performance levels before they pull the trigger on cloud deployments. It seems that if organizations cannot derive rock-solid stability, response times and end-to-end availability from the cloud, they will continue to focus on traditional IT infrastructure, which, as we all know, is known for its rock-solid stability, responsiveness and availability.
This is all the more reason why enterprises need to adopt a multi-vendor cloud strategy, according to Nirvanix' Steve Zivanic. The whole idea of "mastering" the cloud or trying to devise the perfect cloud environment is wrong-headed, he says. Rather, the cloud should be treated like any other IT infrastructure: prone to disruption and sorely in need of backup. It's been a rule-of-thumb for decades not to put all your eggs in a single basket, and there is no reason why that advice should not be heeded just because the new infrastructure is cheaper and more scalable.
This doesn't mean that service from a single provider cannot and should not be improved, according to datacenterknowledge.com's Rich Miller. AWS, for example, provides up to four availability zones per region, plus tools like elastic IP addressing and multi-zone RDS instances to utilize them if one zone goes down. The company provides a range of best-practice guides and white papers at its AWS Architecture Center covering everything from disaster recovery options to building fault-tolerant cloud applications. But as F5's Lori MacVittie notes in the comment section: Relying on a single site/region is like load balancing one server.
So the moral of the AWS outage is that the cloud is not infallible and is not the answer to all your IT dreams. In that way, it's an awful lot like traditional enterprise infrastructure.