Amazon Web Services: The Cautionary Tales

It's been said that you learn more through adversity than good fortune.

By Arthur Cole | Posted Apr 27, 2011
Print ArticleEmail Article
  • Share on Facebook
  • Share on Twitter
  • Share on LinkedIn
It's been said that you learn more through adversity than good fortune.

If that is the case, cloud users can take heart that Amazon's recent outage, although troublesome, particularly now that it seems some customers suffered permanent data loss, was at least a manageable situation that revealed a number of important revelations regarding the right way and wrong way to both deliver and consume cloud services.

In a word, the chief lesson here for users is redundancy. As GigaOM's Derrick Harris points out in a quick round-up of affected customer experiences, those who prepared themselves for the failure of any single host or had robust backup mechanisms in place were able to restore themselves quickly and relatively unscathed. The worst story came from an unnamed home-cardiac monitoring service that couldn't perform its life-saving function for close to a day, although the story didn't explain why such an important service was left wholly reliant on the cloud.

And that's one of the major take-aways here, according to cloud supporters like Dave Jilk, CEO of Standing Cloud. The naysayers, in fact, are right when they say you can't trust the cloud. So don't trust it. At least, don't trust a single cloud provider. By making sure you have ample mirroring across multiple providers, as well as fully automated deployment and management processes, the loss of a single provider shouldn't be any more painful than the loss of a single server.

Of course, this doesn't let Amazon off the hook completely. After all, it gets paid to do what it does, and major failures like this should be followed by a long look inward rather than tearing into its customers' deficiencies. According to Klint Finley on ReadWriteWeb, one thing Amazon might want to examine is the practice of mirroring data across multiple availability zones that are all dependent on a single control plane (aka "point of failure"). When it became overwhelmed, the redundancy that was supposed to exist across zones was an illusion.

At this point, the question of whether the cloud is good or bad is moot. Like it or not, the cloud is here and it will assume just as much functionality over data environments as CIOs deem worthy.

But it would be a mistake to deny yourself the incredible resources the cloud has to offer just because there are still bugs to fix. After all, Thomas Edison no doubt burned his fingers an awful lot before perfecting the light bulb.

Comment and Contribute
(Maximum characters: 1200). You have
characters left.
Get the Latest Scoop with Enterprise Networking Planet Newsletter