There's a palpable generation gap between this new IT culture and old IT, which was on high-powered display during an executive roundtable at VMworld entitled "The Data Center is Dead, Long Live the Data Center."
The panel featured Web pioneer and Silicon Valley venture capitalist, Marc Andreessen, VMware CEO Pat Gelsinger, Sun co-founder and current chief development officer at Arista Andy Bechtolsheim and Graeme Hay, head of architecture and infrastructure for Credit Suisse. Despite his keynote touting the company's software-defined data center vision, roadmap and supporting products, Gelsinger clearly sees the cloud as a secondary adjunct to homegrown and operated enterprise data centers, and was supported in this view by one of his best customers at Credit Suisse. Andreessen was having none of it, and two exchanges highlighted the chasm between world views.
One of the self-congratulatory points Gelsinger made is that virtualization software like vSphere and vCloud have helped IT move the needle on the percentage of time it spends on operations and maintenance versus developing new services and innovating. From a ratio of about 70:30, he presented data showing that, at least among VMware customers, the split is now closer to 60:40, on its way to his stated goal of 50:50. Our survey, analyzed in this year's InformationWeek data center research report, puts the split at 64:36, with the level of innovation increasing four points in the past year.
Unfortunately for Gelsinger, his on-stage customer, Hay, refuted VMware's numbers. When asked, he said the ratio at Credit Suisse was more like 80:20, citing the firm's large and expensive legacy infrastructure and software, along with the overhead and red tape that come with running a financial enterprise.
Andreessen had heard enough, blurting out that at the companies he funds and works with at his VC firm, the numbers are more like 99% innovation, 1% operations. The IT old guard must have thought this was somewhere between naiveté and bravado since in their world, no one could conceivably outsource all their IT--save a few MacBooks and smartphones--to the cloud. But Andreessen was unyielding, reiterating the point that Gelsinger and his co-travelers are setting the bar far too low. Indeed, as one who almost exclusively lives in the cloud--where my primary work machine is a Chromebook, productivity suite is Google Apps for Business, and note and information database is Evernote--I can assure you he wasn't bluffing.
[Read how data centers are becoming the domain of service providers as smaller enterprises increasingly outsource their data center operations in "Data Center Study: The Big Get Bigger."]
The fissure between old and new IT was further illustrated when the discussion turned to the importance, or lack thereof, of internal enterprise data centers, now and in the future. Gelsinger contended that enterprises will be building private data centers for the next 40 or 50 years. The implication is clear: Forget the hype; the cloud will remain a secondary source of IT infrastructure for generations. This is completely at odds with the startup culture Andreessen lives in and ignores the increasing economies of scale accruing to mega center operators, whether commercial hosters like Switch or DuPont Fabros, or cloud services like Amazon, Google and Rackspace.
At this point, I joined the discussion. The moderator had asked the audience to Tweet questions. In light of Gelsinger's comment, I asked, "Why is building your own data center not like building your own power plants?" With only 140 characters I couldn't add any context, and the panelists proceeded to misinterpret my point, saying, well yes, they kind of are. Going to the microphone, I reprised author Nick Carr's argument. Just as companies stopped generating their own power decades ago as the cost of building and operating plants was undercut by the price at which utilities with massive, multitenant plants and intricate distribution grids could deliver the same energy, why would enterprises continue to build their own information plants (data centers) when utilities (cloud services, co-los) using the Internet (grid) can do it far more efficiently?
Once again, the old/new rift was evident. Everyone agreed that smaller data centers are increasingly untenable and that you need scale for facilities to make economic sense. But the old guard, represented by Gelsinger and Hay, clung to the notion that on-premise, self-managed data centers maintain sufficient--what I would contend are intangible or illusory--values of control, security and operational responsiveness and accountability to justify their existence. Aside from the fact that most enterprise data centers today are tiny--almost 72% of InformationWeek's survey respondents operate facilities smaller than 10,000 square feet--these "values" aren't demonstrable.
As Andreessen pointed out, most clouds are more secure and reliable than enterprise infrastructure. Gelsinger countered by citing the recent Amazon outage. Again, Andreessen had a response. It's the classic "man bites dog" news story: The Nasdaq exchange goes down for more than three hours and it's just the price of complexity in our technological era; a fraction of Gmail users see a glitch for a few minutes and it's proof of the cloud's fragility.
Like all generation gaps, there's no bridging old and new IT; the old just fades away. Old IT will cling to its data centers, on-premise infrastructure and self-managed applications until the weight of inefficiency and inflexibility swamps their businesses under a tide of nimble newcomers using utility resources, SaaS applications and rapidly adaptive strategies. SoMa and Silicon Valley are full of such companies run by brilliant risk takers and advised by visionaries like Andreessen and his Sand Hill Road colleagues. Big enterprise IT needs to get out of its bubble and see this rapidly growing new world before it's forced into an early retirement.