Why enterprise CIOs shouldn't be worried about the AWS outage

Last week's much-discussed Amazon Web Services (AWS) outage reignited simplistic debate about whether public cloud computing was "ready" for widespread use.  The usual naysayers jumped like bullies on a perceived sign of weakness from the leading infrastructure-as-a-service (IaaS) provider, who happens to also be the leading example of the innovation made possible by this new model.  Shamefully, some alleged advocates of cloud computing for the enterprise joined the fray with it's-time-to-panic soundbites like, "if you put all of your eggs in one basket, you put yourself at risk," in directing blame to Amazon's customers for their own troubles.

This kind of "I told you so" finger-wagging is ill-timed, particularly with regard to the AWS incident's implications for the enterprise, for several reasons:

• Amazon's infrastructure offering does, in fact, offer better levels of backup, failover, and load balancing than most departmental IT teams are prepared to develop for internal applications. Actually building a full-fledged high-availability cluster is too costly, and adds too much complexity, to the typical enterprise application.  Software architects are, and always have been, obliged to weigh downside risk of outages against these costs and complexities. Phil Wainewright quoted O'Reilly's Jim Reese on this point: you "either deemed an outage of this nature an acceptable risk" - as it would be, frankly, for most enterprise apps - or you design a more failsafe system, which can be done using either on-premise or, yes, on-demand platforms.
• Most of the post-snafu attention was aimed at outages among consumer-oriented websites.  The direct productivity impact among enterprise users of AWS was likely lower, with more batch-oriented processing and other forms of distributed workloads with less emphasis on serving up end-user web pages.
• By highlighting the impact of a single large outage, we obscure the fact that countless small outages were avoided by the very use of AWS.  Rackspace's Lew Moorman illustrated this with a clever analogy, calling the incident the "computing equivalent of an airplane crash," which is a major episode with widespread damage, but doesn't change the fact that airline travel is still safer than traveling in a car.  The cost of a great number of small outages that occur every day behind closed firewalls easily outweighs the impact of this one event that happened to impact many users simultaneously.

Consolidating the world's computing power, and the sophisticated controls required to harness this power, is an economic inevitability.  We are, inexorably, advancing towards a day when most CPU cycles consumed by enterprises are executed on public cloud infrastructure.  But this doesn't mean we are all moving overnight.  And the fact that our transition will be gradual doesn't diminish the credibility of those who have already taken the first steps.  To read some AWS critics this weekend, you'd think the "cloud first" CIOs had been proven wrong.

They haven't.  "Cloud first" doesn't (yet) mean "cloud only."  The reverse-named ComputerWorld opinion piece "Why You Don't Need a Cloud Computing Strategy" (coincidentally published three days before the AWS outage), in making its case for why you do need such a strategy, includes these salient points:

• "Rather than trying to replace existing infrastructure that is already working, identify workloads that are dynamic or new"
• "Think small, but plan big"
• "Acknowledge the immaturity of cloud computing, but don't let it hold you back"

This mirrors advice I've given frequently to clients while formulating a cloud computing strategy:  You don't need to rip out all of your working software.  But you should start considering capping your current level of investment in on-premise infrastructure and systems.  In other words, don't throw good money after bad.

The New York Times recently published piece that looks back at how quickly revolutionary innovations have resulted in widespread productivity gains.  "By 1987, the personal computer revolution was more than a decade old, when Robert M. Solow, an economist and Nobel laureate, dryly observed, 'You can see the computer age everywhere but in the productivity statistics.'" The piece asserts that it wasn't until 1995 that personal computing had a significant impact on the American economy - roughly a 20 year lag.  Any takers willing to bet that cloud computing moves more quickly up the value chain?

Glenn Weinstein is the CTO and co-founder of Appirio, where he oversees the CloudWorks and Cloud Management Center product lines as well as internal IT.