Amazon's Storm Clouds

Whether it’s your data center or someone else’s, downtimes still happen

Amazon EC2  is a cloud computing service operating from multiple data centers around the world, including at least Singapore, Tokyo, Ireland, and in the US, California and Virginia.  Businesses can have servers operating on their behalf from one or more of these locations. Amazon speaks of more conceptual groupings of facilities with terms such as “availability zones,” but ultimately collections of physical and virtual servers are running in physical data centers.  Those data centers are part of the real world and can still have real world downtime.

Amazon has a status page online and that page showed a pretty rough couple of days recently for one of the locations where Amazon operates its EC2 service.  Those who had services running from that location may have experienced a serious disruption for an extended period of time.  If a service was running at only that location, it was very likely down.  Popular services like reddit and foursquare suffered as a result.  Many lesser known businesses did as well.

Should this steer us away from Cloud Computing?  Certainly not. The Wall Street Journal correctly pointed out that Amazon does not guarantee 100% uptime, and that it’s possible to use Amazon’s services (or any combination of servers) to protect against single failures.

In ZDNet's Seven lessons to learn from Amazon's outage, the author points to some of the key responsibilities of the enterprise to understand the cloud service and how to use it carefully and effectively.

If businesses think twice in the near future rather than simply assuming that cloud computing is a silver bullet of low cost, flexibility and resiliency, that will be a fine outcome.  I think the single lesson is that running “Always Available” online services isn’t a matter of enterprise data center versus cloud services, and it isn’t a matter of which big vendor’s software you run.  It’s a matter of careful design by qualified IT professionals who know how to make best use of those enterprise data centers, cloud services, and software (vendor proprietary and open source alike).  It involves serious technical savvy and an understanding of business objectives.  Just like it did before we put services in clouds.

Please leave a comment about Cloud Computing, about designing for availability, or anything else!  And please follow me on Twitter as @dkassabian.

Links:

Amazon Service Health Dashboard

http://status.aws.amazon.com/.

Wall Street Journal

http://onespot.wsj.com/small-business/2011/04/25/b299b/amazoncom-goes-down-is-online-software

ZDNet

http://www.zdnet.com/blog/saas/seven-lessons-to-learn-from-amazons-outage/1296