Outages: growing pains at salesforce.com

Salesforce.com (SFDC) has recently been getting the wrong kind of attention about its outages - Salesforce.com Crashes Again. There's a chorus of people complaining in public - company watchers, analysts, customers. Competitors are opportunistically picking on it as well. Wall Street initially got a bit nervous and slammed SFDC a bit, although they seem to have come to terms with it. And check this out - a blog called GripeForce!

So what is the ballyhoo about?

At a perception level, its simply people asking a poster child to be more accountable about, well, being a poster child. Outages like these, even more, are common with on-premise software. But on-demand outages occur at a concentrated point of failure, affect lots more people simultaeously, and this sets up a big common target whose neck can be wrung. The market even has an incentive to talk more publicly about it than it has for on-premise failures.

At a service operations level, outage risk is inherent with shared infrastructure and simply needs the right kind of risk management practice - usually, the right level of investment in infrastructure redundancy and monitoring tools, with an SLA providing the sort of skin-in-the-game that gives customers the assurance they need. And heck, outages are like teenage awkwardness...everyone that grows up in the web world goes through it. Yahoo had them back in 2000, Ebay back in 2001, and Google got hit last year.

From SFDC's standpoint, their outages are actually not a bad problem to have. Sure they've probably gotten a bit ahead of themselves with the explosive growth they're seeing, the new products (err...services) they're building and the market foment they're working up. They did see some of this coming and started taking steps to avert it a few months back with the MirrorForce datacenter investments. I'm sure they also have more coming to address Forrester's recent comment “Salesforce.com: Time For A Standard SLA”. I would'nt doubt this company's ability to execute on growth.

And while competitor RightNow was certainly not as strained about managing growth as SFDC, they do appear to have made very different operational choices. With due credit to them, they're now even crowing about it. A couple of little secrets about RightNow's sauce is that they don't actually run their own data center...they use co-located facilities, and they run their platform on open-source tech components. To some extent, RightNow's uptime performance validates a couple of very useful things to know about running on-demand operations - 1) that data centers are best run by specialists that manage large scale infrastructure...a point that Nicholas Carr makes in his post 'Salesforce.com's hiccups' . 2) Open-source tech components have a high enough degree of reliability to run SLA grade on-demand operations and if there's a glitch..Can We Fix It...Yes We Can! (Just got carried away a bit there with my son's Bob-the-Builder slogans...). So what does SFDC do when something like this happens - they point their finger at their proprietary database vendor. Tough luck. Tough choices.

Getting past the operational level, matters start to get really interesting from a market strategy perspective around outages. What is a fair and reasonable benchmark for application availability levels? Is 99.98% good enough? Are there customers who can manage their application infrastructure better on their own? Are they a useful market for on-demand vendors to address? How much better can those customers manage and what is it worth to them to do it? How much are they willing to spend doing it? Is it worth segmenting and serving market demand along varying degrees of SLA availability levels priced differently per customer needs? Or are multi-tenant on-demand platforms like economy-class air travel - inherently self-segmenting in terms of the market segment they address and the mission criticality of customer needs that they serve. Basically goes like this - if you're one of the sorts who needs home-in-the-air sort of air travel, whenever you like and to wherever you like, you don't buy an airline ticket. You buy and run your own corporate jet. ;)

So while its fair for customers to demand benchmark service availability levels and for vendors to bear the cost of delivering it, there may actually be some opportunity around outages as well.