Almost the last word on salesforce.com's outage

Speaking today at the Merrill Lynch IT Services and Software conference in Las Vegas, Salesforce.com CEO Marc Benioff didn't directly address the outage that occured earlier this week. He said that over the last seven years, salesforce.com has had excellent reliability, and has never fallen below 99 percent. Even with the outage in December 2005, uptime was 99.7 percent, he said.  He attributed the two outages to shaking out new data centers.

As a reminder about uptime:

99.5% — 43.76 hours (an entire working week, and more)
99.7% — 26.30 hours (more than three working days)
99.9% —  8.77 hours (more than one working day)

(see Phil Wainewright's post on uptime)

I sent salesforce.com another request after posting about the company's lack of specificity regarding the January 30 outage, asking for details on what a subset of customers view as more than a minor inconvenience. (Phil and David Berlind blogged about the salesforce.com's lack of transparency as well today.)

I also asked whether the company's reliance on large Sun servers that can handle hundreds or thousands of customers versus an architecture based on commodity servers that each support less than a hundred customers is an issue. Deploying new software in more scale-out, commodity server approach would affect fewer customers at once. Phil also probes that question based on comments from competitor NetSuite's CEO Zach Nelson.  Nick Carr talks about the issue in terms of infrastructure and application service providers as distinct areas of expertise here.

Bruce Francis, vice president of corporate stategy responded as follows to my query for more info:

The big picture is that over the past seven years we have an excellent track record of service.  And we never stop working at improving it.  The big picture is that our service allows our customers to achieve greater uptime than they could buy and maintain on their own—and at far lower cost.

That sounds a lot like Scott McClellan at a White House press briefing avoiding the questions. Salesforce.com executives say, go talk to customers. We did, and it doesn't fully jibe with the company line. One large customer said that although the Web application (search, reports, etc.) was available, the salesforce.com API was down for about six hours on January 30. This meant no access to the Outlook or Offline editions or partner products from AppExchange that require the API.  

Clearly, salesforce.com doesn't believe that it is obligated to publically disclose specific data about the outage. By not engaging in a more transparent dialog, the company is risking more serious backlash the next time it encounters a reliability problem.

The net result of all this conversation about salesforce.com's uptime--as the highly visible representative of the new class of hosted software and infrastructure providers--may be that service level agreements become more than just an instrument for the largest customers, and that the software-as-a-service providers will have to put some money away for paybacks when service levels aren't met.

Infoworld has a story about salesforce.com customers clamoring for more information and SLAs. Computerworld weighs in with comments from salesforce.com customers who say that the advantages of the salesforce.com platform outweight the glitches. One customer quoted, with 125 users, said that the users find something else to do when the system is unavailable and that salesforce.com is more reliable that their internal systems. Not all customers are created equal. For some, salesforce.com is mission critical, which carries a different set of requirements.