Sorry, The Internets Are Broken Today

This past month has not been great for cloud computing, and yesterday was no exception. First, I try to visit Microsoft's MSDN forums to follow up on a question, and found that they were "down for maintenance" most of the day. Then, GMail has a two-hour outage in the afternoon.One of the great things about Web-based services is that they can be fixed quickly. In the past, security bugs discovered in Google pages have been fixed the same day. This past month shows the other side of that coin; Web-based services can break just as quickly, and stay broken for hours. The GoToMeeting, SiteMeter, and Amazon S3 outages certainly didn't help the case for cloud computing.

Computers certainly can fail; the question for a Web site administrator is which failures to anticipate and how to deal with them. The more third-party services that you use, the more your site's reliability and security is out of your hands. It's the same control-versus-cost decision that leads companies to outsourcing -- and sometimes leads them back.

Sites definitely should have contingency plans in the event of an outage, but third-party services make this even more difficult to achieve. In the case of SiteMeter, the problems that were occurring couldn't be detected at the Web server's end; error messages were being shown to clients but everything was fine as far as the Web server was concerned. Due to the problem happening late on Friday, and the nebulous nature of IE's error message, it was difficult for most sites to realize that their third-party Web stats were to blame.

That's the contrast with traditional infrastructure. When you're delivering every bit of content on the page, it's a lot easier to track, manage, and secure it. When you use a third-party service, it becomes a black box that is hard to verify, or even know if or when something has changed. Welcome to your future nightmare.