3 questions to consider before moving a data center to the cloud
The old ways of running data centers won’t last. Here’s what the future might hold.
- By Brian Robinson
- Sep 24, 2009
In the old days, many information technology departments had an elitist air. Their specially trained staff and expensive machines created the magic behind the curtain that helped other departments in the organization get their work done quickly and more efficiently. So it was always a good idea to tread lightly — and bring treats — when it came time to ask IT for help.
However, newly empowered users aren't so deferential to the denizens of the data center. Time-saving applications and ridiculously powerful computers that fit a dozen to a shopping cart are available with a mouse click or at the office supply store down the street. So why put up with the attitude from IT?
For many good reasons, not the least of which are security, efficiency and, yes, expertise, IT departments remain central to almost any business or government enterprise. That’s not to say there aren’t some problems with doing business as usual. Most data centers are still not always able to give people what they want or need as rapid technology innovation creates change and new opportunities.
How can the IT department continue to be the go-to player in this new technology era?
That question has prompted many chief information officers to start thinking about cloud computing, but not the cloud of recent industry hype and speculation, in which commercial firms provide access to software applications and computing platforms via the Internet for a fee. That will come later, experts say.
Rather, government technology executives, including some in the Defense and Energy departments, are thinking about or already remaking their own data centers in the image of a cloud. They are delivering IT services to their internal customers with operations that are efficient, flexible and fast.
And those adjectives are more than just buzzwords. CIOs know that those qualities can translate into significant cost savings, both for the data center and for the departments they serve.
Here are three key questions agencies are pondering as they consider that kind of move.
1. Why should agencies set up an internal data center like a cloud?
Traditional data centers are all about certainty. They contain a large level of redundancy and extra capacity for servers, power supplies, storage and networks so the applications — and their data — hosted on the infrastructure are always available to users.
However, the penalties for building such an infrastructure include system sprawl, management complexity and ever-increasing power demands -- all of which inflate costs.
“The data center is the most costly building in the federal government,” said Peter Tseronis, DOE's deputy associate CIO. The department is exploring cloud concepts as a way to lower costs. That makes data centers “the perfect low-hanging fruit for making improvements.”
One way the cloud can drive cost out of the system is through server virtualization, in which single servers are partitioned into a number of virtual machines, each of which can run its own instance of an operating system and applications. Those virtual machines can serve different customers but all reside on the same physical server. That enables agencies to consolidate servers in the data center, which reduces costs.
It’s also difficult to quickly meet changing user demand in a traditional data center, where someone needs to manually configure each new application or workload on dedicated computing resources that must be reorganized or bought. In the cloud model, applications are standard services that administrators can call on when and how users want them, with little or no regard to the particular computer they are running on.
With the flexibility to share resources, administrators might be able to get away from the traditional practice of budgeting for capacity that meets the high-water mark of each workload.
“Most of the resources in dedicated data centers are only fully utilized at peak times,” said Venkatapathi Puvvada, vice president and managing partner at Unisys’s federal horizontal service segment.
And more than one agency might be able to share a flexible environment, thereby increasing overall efficiency.
“Over time, it will make it much easier for agencies to share data and applications across the enterprise,” said Clint Harder, sales and product development manager for advanced technology services at CDW Government.
2. What do agencies need to do to build a cloud?
The first step is to virtualize the server environment by creating and following a virtualization strategy. With the strategy for virtualizing servers in hand, administrators can evaluate how to apply it to other parts of the IT infrastructure.
“The impact of virtualization is not just at the machine level,” said Jeff Bergeron, chief technologist for the U.S. Public Sector at EDS, a Hewlett-Packard company. “You need to look at both systems and networking, as bandwidth will have to increase in order to handle users’ self-provisioning of services.”
Administrators also need to redesign storage systems to accommodate a fluctuating and increased data flow. And a broader use of blade server technology will be important.
Blade servers shrink the power of a PC-type server onto a slim card that fits inside a standard computer chassis to provide more processing muscle. They also typically use less power than standard servers, so they can help a cloud-like data center begin to reduce energy costs.
System automation tools are necessary to enable users to provision services so they don't need to ask IT managers to handle that task. Load balancers are another important ingredient in the cloud model to make sure that systems efficiently handle shifting workloads, said Rue Moody, strategic products technical director at Citrix Systems.
Meanwhile, most agencies have started investing in those kinds of tools and technologies and are familiar with their operation.
3. What obstacles prevent agencies from moving to a cloud?
Security is the biggest barrier for agencies interested in moving data centers to the cloud. It is a concern in traditional data centers, but at least software applications and data reside on known servers. Knowing where to apply security in that model is much easier.
On the other hand, the cloud is a multitenant environment in which multiple servers and storage devices can contain a single department's data. Likewise, a single server or storage device could host data from several agencies.
A risk analysis can help managers determine which applications are appropriate for a cloud environment and which are too sensitive and should remain on dedicated systems. Managers need to pay attention to interoperability issues for existing applications that were previously isolated but might now interact in the cloud, Puvvada said.
Many government IT departments use service-level agreements with some internal customers to define performance requirements, such as system availability or application response times. Such customers often have dedicated systems at the data center that serve only their needs because their SLA requirements are so rigorous.
Because the cloud model is a shared infrastructure, SLAs become a necessity, said Doug Chabot, vice president and principal solutions architect at QinetiQ North America. IT staff members will use SLAs to manage workloads across systems. Internal customers will use them to monitor the performance of cloud services.
“These are the kind of front-end issues that have to be tackled,” Chabot said.
Finally, any move to a cloud model will need to deal with the existing mindset. At many agencies, groups that produce data have traditionally owned it. That attitude won’t fly in the cloud.
“If this culture of ownership persists in the data center, then cloud computing will not be fully realized," said Deniece Peterson, principal analyst at Input's industrial analysis group.
In the end, Tseronis said, agencies' biggest barrier to moving a data center to the cloud is the level of risk that they are willing to assume.
Brian Robinson is a freelance writer based in Portland, Ore.