According to a Symantec survey, discussed in NetworkWorld, half of all data centers are understaffed: 34 percent are “somewhat understaffed” and 16 percent are
“severely understaffed.” One poignant quote came from a network manager who said “We recently actually lost 25 percent of our departments in
cutbacks. I’m now doing a number of different jobs. I work from home and on the weekends doing things I never had to do before. I am just trying to do
more with less. We could use more people, but right now, it doesn’t seem to be an option.”
Slideshow: Secrets of Successful Data Centers
Data Center Definitions and Solutions
As I interpret his statement, he’s not actually a manager at this point, he’s a sysadmin—or at least he’s a manager who has taken on sysadmin
duties as well. Overall, the message from the survey is that IT executives are squeezing costs out of their data centers—although, to be fair, 45
percent of survey respondents said their organization is sufficiently staffed and another 5 percent said they are overstaffed.
Also noted in the survey is how challenging the cost pressure because “data centers are becoming more and harder to manage, with more
applications, data and increasingly demanding service-level agreements.”
This is an IT perfect storm: higher requirements, fewer resources. And, with all due respect, the aphorism “do more with less” has always struck me
as beyond the border between fortitude and fiction.
The reason for this manager/sysadmin’s (or should it be sysadmin/manager?) burden is clear: IT is being treated as a cost center, and in tough times
cost center budgets get chopped. Gartner’s just-released IT budget
report says budgets will increase 1.3 percent in 2010, compared to 2009, when they fell 8.1 percent. In other words, IT budgets will be around 6
percent lower than they were in 2008.
And yet, we hear more and more about how the real move to cloud computing will be private clouds—IT organizations implementing cloud
computing services in their own data centers. The reasons for this vary, including concerns about security in public clouds, putative cost savings from an
internal service group unmotivated by profits, the domain knowledge brought to bear by an internal group versus an outside provider, and so on.
The Case For and Against Private Clouds
Does this cost-cutting square with IT organizations building out their own private clouds? One has to question whether the two things can both be
Because if there’s one thing that you can count on, implementing a private cloud won’t be cheap. Here are the areas that require investment:
1. Hardware: To implement a fully agile data center, all of the hardware must be virtualized—servers, network, and storage. This
means latest generation kit, as it is only in the past year or two that vendors have upgraded their offerings to be fully virtualization-capable.
2. Software: Obviously, the latest generation virtualization software must be implemented across all the hardware that will comprise the
3. Human Capital: Learning how to run all of the new kit, manage the virtualized resources, and operate an agile data center requires new
skills that will require training and time to come up to speed. This upgrade to human capital will—at least temporarily—reduce productivity,
which in a headcount-starved IT organization seems perhaps challenging to implement.
It seems that upgrading even a modest-sized data center will cost millions of dollars —and keep in mind that the full investment is required for
whatever portion of the data center is turned into a cloud. In other words, unless you spend in all three areas at the level required to make them
cloud-ready, you don’t have cloud at all. It’s an all-or-nothing proposition.
Of course, the incumbent vendors to IT organizations aren’t going to discourage the notion that an internal cloud is just the ticket. With so much
capital spend potential, they have a vested interest in minimizing the reality of how much work and money are going to be required. As Werner Vogels,
CTO of Amazon recently tweeted: “thinking about starting public wall-of-shame website for sales VPs who promise to help convert my IT infrastructure
into a private cloud.”
This is not to say that private clouds won’t be built. It’s just that the dividing line between those who actually end up implementing one and those who
don’t will be much higher on the “size of IT organization” scale than most people think. Most IT organizations, when they really think through the
implications of putting up their own cloud, will recognize it’s not in the cards for the following reasons:
• Cost Before Value: The capital spend is front-loaded before the benefits of the cloud are realized. If you decide to “cloudify” 100
machines, you’ve got to spend 100 machines-worth of cost before cloud app one gets spun up. In a spend-constrained environment, this doesn’t seem
• Destabilizing Operational Environments: The first rule of IT is don’t mess with something that’s working unless you have to. Upending
a lot of infrastructure to cloudify it when most of the incumbent apps don’t take advantage of cloud characteristics like scalability and agility seems …
dangerous. The inclination will be to leave already-existing infrastructure undisturbed.
• Change Management: Implementing cloud computing requires process re-engineering for how IT operates. Boy, that sounds like fun
and a recipe for high productivity. Imposing organizational change in a personnel shortage situation seems foolhardy.
A much more likely scenario is the following:
• Very large companies that really extract value from IT will implement private clouds. Smaller companies (and these will still be relatively large
companies — remember, the dividing line of where a private cloud can be economically justified is quite high), once the economics of private clouds
become more evident, will lose their enthusiasm for implementing their own. Eventually, most companies that want a private cloud will turn to a service
provider that offers cloud capability. In this sense, we’ve seen the high water mark for internal data center capacity; going forward will be a gradual
decline in internal data center square footage.
• Companies will recognize that one area that can extract value from a private cloud is development and test, where rapid system spin up,
scalability, and agility are extremely valuable. They will carve out a portion of their data center to become a dev/test private cloud. However, IT
organizations may find that developers are too impatient to wait for the time that the capital budgeting and infrastructure implementation will take. I was on
a panel recently in which a very senior executive of one of the largest software companies around lamented the fact that his developers were turning to
Amazon to do their jobs — despite being cautioned about what in his view were intellectual property risks. For developers, productivity and convenience
trump company policy. I recently had coffee with James Staten of Forrester; he is quite skeptical of the prospects for private clouds, but thinks dev/test is
one place they make sense.
• Business groups within companies (e.g., marketing) will get frustrated at the lengthy evaluation, budgeting, etc., etc. pace associated with “IT’s
private cloud” and start implementing their own applications in public cloud providers that offer pay-by-the-drink pricing. This phenomenon, often
denigrated (by IT) as “shadow IT” is going to be powerful. In the not-so-distant past, business groups had to come to IT eventually to get their
applications running on metal. Today, a company credit card and an Amazon account enable completely bypassing IT. Note that this phenomenon may go
on even after the IT organization has set up the official “service provider cloud.” Absent re-engineering and cost-competitiveness with public clouds,
business groups may still choose to use the easily accessible and cheaper alternatives available to them.
Despite the hoopla about IT organizations building their own private clouds, it’s hard to see how constant cost pressure will support the investment
necessary to implement them, notwithstanding the hypothetical cost savings a private cloud will ultimately bring. Much more likely is a rapid move of
greenfield application implementation to cloud service providers like Terremark or to public clouds like Amazon, with a gradual migration of existing
applications to those environments as a result of hardware obsolescence or software refresh. The challenge for IT organizations will be to overcome their
“NIH” (Not Implemented Here) attitude about cloud computing.
Bernard Golden is CEO of consulting firm HyperStratus, which specializes in
virtualization, cloud computing and related issues. He is also the author of “Virtualization for Dummies,” the best-selling book on virtualization to
Follow Bernard Golden on Twitter @bernardgolden. Follow everything
from CIO.com on Twitter @CIOonline