Charlie Wood is like a pioneer trekking across the frontier in a covered wagon, but the dangers aren't an untamed land, broken wheels or hostiles. Charlie, who runs the startup Spanning Partners, which helps companies RSS-enable applications, is on the frontier of using the software-as-a-service, utility computing model to run his business and is coming face-to-face with the unreliability of the on demand services he depends upon. In his blog, Charlie writes:
Yesterday Amazon's S3 service suffered an outage that lasted more than long enough to miss the company's self-imposed goal of 99.99% availability, at least for the next couple of months. Last November, Google Calendar was unavailable for the better part of a day. And famously, Salesforce.com suffered a string of outages this time last year. Unfortunately, my business relies on all three of these services.Certainly, the higher cost of conventional hardware and software solutions is consideration, and those systems are not immune to failures and excessive bloating. But the SaaS, utility computing model has prove that it can be more reliable, responsive, secure, lean and cost effective to replace what came before...
We're clearly still in the very early days of utility computing, but reliability is an issue that will have to be addressed sooner or later. Ultimately, it comes down to accountability. Bad things will happen. The question is how the vendor will respond.
When Amazon's S3 service tanked, an Amazon Web Services engineer posted a message to a discussion group with an acknowledgment and a lame apology. When Google Calendar went south, they gave much the same response. When Salesforce.com fell over, they spent $50M and an entire release cycle fixing the problems.