It being McKinsey, you can assume the math is correct. For once, though, I think the numbers fail to tell the story. Though whether that’s their fault or McKinsey’s is open to question. But I’m getting ahead of myself.
For those that missed it, the folks over the Uptime Institute engaged the celebrated management consultancy in a project to look at cloud computing. After examining the technology and the economics, McKinsey produced a presentation entitled “Clearing the air on cloud computing” that concluded that while cloud computing has undeniable benefits, “many of the claims being made about cloud computing have lead some to the point of ‘irrational exuberance’ and unrealistic expectations.” In Uptime’s words, anyway. The New York Times was a bit more blunt, saying “But while cloud computing is a marketing triumph, new research from McKinsey & Company asserts that trying to adopt the cloud model would be a money-losing mistake for most large corporations.” Further, they quote the study’s author, Will Forrest, as follows: “The industry has assumed the financial benefits of cloud computing and, in our view, that’s a faulty assumption.”
Now it’s true that saying that the technology industry is fickle is something of an understatement, obsessed as we all are with the shiny new thing. And with the task of marketing software a difficult task in the best of times, and these being very far from the best of times, it’s as unsurprising as it is predictable that the term cloud computing is being thrown around like it was, say, SOA. So the conclusion that cloud computing is just the latest marketing fad might seem logical. Inevitable, even.
But like Tim, I think that belies the real inevitability, which is the heavy adoption of cloud offerings.
The fact that the economics, apples to apples, are not a landslide win for the cloud would not be – or should not be, anyway – a surprise to many. For $69 a month, you can get a server from 1and1 that’s almost twice as fast (2.2 Ghz vs 1.2 Ghz) with twice as much physical memory (2×160 GB v 160 GB) as the default Amazon EC2 instance, which would run you approximately $74/mo. McKinsey’s delta between the loaded total cost of a CPU managed by a datacenter and EC2 are even more damning; the former beats the latter $150 to $366, for a 144% delta. Ugly.
Or it would be, if it were apples to apples. But it’s not.
Tim’s primary criticism of the McKinsey piece is that economics are not the sole predictor of success within the technology space.
I can remember like yesterday sometime in the Eighties, the CIO of my parent company, a telecom-equipment manufacturer, lecturing executives about how we simply could not realistically think about meeting information-worker demand for PCs, and I’m pretty sure one of her slides (a real slide, on a real projector—you had to allow a week’s lead-time to prepare them) had more or less exactly McKinsey’s points. She lost, because the information workers simply routed around her.
Getting Stuff Done · At the end of the day, that’s all that matters. When a new technology arrives, if it’s basically affordable and it lets you do your job faster and better, well, it just doesn’t matter if it’s a little more expensive or a little less secure or any of those other things. It’s going to happen.
Precisely. McKinsey’s numbers, in my opinion, are being handed down from an Ivory Tower. Similar to studies that “prove” how much more productive, maintainable or secure Java is to PHP, it ignores the most important feature of cloud computing: a low barrier to entry.
While having lunch in New York a few months back, I had the opportunity to talk about EC2 with some of the developers from a large media organization. They referred to it as “infrastructure that you can buy as books.” With budgetary limitations, their ability to procure more hardware had been severely curtailed. Working at a meda company, their ability to purchase books had not. And while such infrastructure-as-books anecdotes are flip, they are simultaneously anything but. The fact is that whatever the theoretical economics might be for any given environment, the practical market looks considerably different.
Having fought, as a systems integrator, through the procurement, hardening and deployment processes for new hardware for projects, I am highly skeptical that McKinsey’s math – while undoubtedly quite accurate – will trump the convenience of infrastructure on demand. As open source demonstrates quite convincingly, availability and convenience are powerful incentives indeed.
But more than glossing over ease of adoption as a fundamental driver of adoption, I’m perplexed by the implicit assumptions that cloud computing is a zero sum game. Slide 24, for example, assumes the “migration of total Windows and non-console Linux capacity for entire data center,” and slide 25 discusses the “total infrastructure labor base saving potential if moving whole data center to cloud” – they project it at 10-15%, in case you’re curious.
Who would contemplate the migration of their entire datacenter investments to the cloud? Forget the economics: unless the cloud offerings were effectively free – which McKinsey has convincingly proven they are not – the risks of migration would far outweigh the potential economic returns. In a world where US tax returns are being processed by COBOL programs written in the 60′s, I find the notion of wholesale migration – one of the assumptions for portions of the presentation – exceedingly problematic. It’s akin to questioning the economics of converting the entirety of one’s application portfolio to Visual Basic because that language is easy to learn.
Far more likely than an abrupt private datacenter to cloud shift is an adoption pattern similar to what we see today; James is exactly right, in my view, when he says that Amazon is the new VMware. In other words, we can expect pilot projects to be followed by discrete, tactical employment of the technology, bridging in time to strategic but complementary usage. And the economics of that complementary consumption – particularly in the speed of availability and time to market, and later, the impacts on pricing – would change considerably.
In 1980, McKinsey was asked by AT&T to predict the cellphone market in 2000 and was off by a factor of 120. I don’t know that they’re off at all in their numbers here (though their lack of attention to the economic advantages in deployment speed is troubling), but neither am I remotely convinced that they’ll be determinative.
Should enterprises embrace virtualization and “datacenter best practices” as McKinsey recommends? Absolutely. But those that would fight the rising tide of cloud adoption are likely to fare no better than Cnut.
Update: While I was willing to give McKinsey the benefit of the doubt on their numbers, Vinnie Mirchandani wasn’t. Interesting read.