There could be money hiding in your data center labs.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Many of these environments used for testing and development are overbuilt and inefficient. The top challenge related to a data center lab is the time required to manage and maintain it (41%), followed by the cost of the infrastructure (39%), the physical space requirements (34%), and the time required to configure and deploy the lab (33%), according to a recent survey from Spiceworks Inc., an Austin, Texas-based IT community that also conducts online surveys about IT buying habits.
Many enterprises have a test and development lab unnecessarily close to their offices, in space that is either inappropriate or overly expensive for a data center or server room that doesn't run production applications.
"We call it server hugging," said Pete Sclafani, COO and co-founder at 6connect Inc., in San Francisco, who has more than a decade of experience building and deploying data centers.
The 2015 study of 151 IT pros based in the U.S., called Hidden Challenges of Data Center Lab Management, was commissioned by Vantage Data Centers, a data center operator based in Santa Clara, Calif., which sought data about how much overbuilding and overpaying is associated with test and development data centers.
Of course, Vantage has a vested interested in proving that data center outsourcing is a better bet than running a lab in-house, but there are a number of ways to cut corners and reduce costs in either scenario.
Where to cut corners in data center labs
Reducing infrastructure for the test and development data center may not be for all organizations, said Chris Yetman, COO at Vantage and former vice president of infrastructure operations at Amazon.
"It takes the bold to think about it that way," he said, noting that Vantage has one customer whose test environment runs without an uninterruptible power supply (UPS).
If the power goes out, workers are simply told to work on their laptops, check email or go home, he said. The interruption wouldn't affect customers or end users, but some organizations may see it as a bite into productivity.
Location and power draw drives much of the cost of a data center. There's a premium cost to having your test labs in expensive real estate like San Francisco, whereas a move to Dallas or Denver, or a location in Arizona, could save 20%, Sclafani said.
Paying close attention to power rates can also yield significant savings, Sclafani said. In California, Pacific Gas and Electric Company (PG&E) utility power costs more than a renewable source of local power from a supplier, Sclafani said. Silicon Valley Power (SVP) in Santa Clara is 18% to 42% less expensive, depending on the class of service, which is based on rates anywhere from 3 cents to 8 cents per kilowatt-hour cheaper than PG&E, according to SVP.
Eliminating redundant power supplies is another way to save money, but Sclafani said he hasn't seen many examples of it.
"If it is truly a test environment and you don't care if the power goes out, that is probably a possibility," Sclafani said.
By reducing their requirements for things such as UPSes and generators, companies could cut costs by 15-20%, he said.
But if a nonredundant, less-than-resilient environment is all that's needed for a data center lab, wouldn't an office setting be just about right? For smaller test environments -- for example, in the 100-200 kW range -- an office or server closet works fine, Yetman said.
Though the office may have enough power for that size environment, office buildings aren't designed for racks of servers.
"Even if you are willing to take a risk with power, you will still have to deal with cooling, even if you are willing to run it warmer," Yetman said.
An office building may have power usage effectiveness of 2.0, while a data center could get the PUE down to 1.3, he said. Plus, even with proper cooling, a decision to run the test environment warmer could make sense. "Have it be a T-shirt and shorts data center," Yetman said.
One other option to move the test and development data center from the office is to rent servers from a dedicated hosting service. This is great for the development process, Sclafani said, because it gives organizations the ability to "try before you buy."
An organization may develop an application on its own infrastructure, but wants to test its performance on Intel Xeon processors, for example, to understand if new equipment would increase performance. This approach can also help to determine the amount and type of storage needed for the application.
"You just want to get some benchmarks," Sclafani said.
A complete move to cloud computing for test environments has also become quite common. Often, the test and development environment becomes the production system the application will be in during some point of the development cycle, so VMs or the cloud can be a great way to do basic functional testing, Sclafini said.
For example, if the plan is to have the application go into production in a cloud environment, IT pros do functional testing and development on any combination of the cloud, VMs and dedicated servers. That helps right-size your cloud for your application and helps confirm the projected cost to run the application.
Another money-saving technique is to request a lower-level network connection, often known as the "blended transit option." Its equivalent of the house wine, Sclafani said. Then, as a backup, get a contract with a top provider with a low commitment.
Also, don't overlook your organization's service-level agreements for support, Sclafani said. Maybe 24-hour daily support isn't needed for data center labs as it is in the production environment.
Don't pinch pennies on security
There are all sorts of corner-cutting options to save money on a test and development data center, but physical security should remain top notch, Yetman said.
He has personally been "ripped off," and said organizations should be worried about the threat posed by someone showing up in a uniform and work belt to gain access.
"There's espionage out there that happens like that," Yetman said.
In many cases, managing the network and making sure it is secure should still be job number one for security.
Yetman pointed to a recent example of a Seattle company that lost data when someone walked into an office building where a test and development data center is located, grabbed a development server and walked off. Physical security should be assured, so data center managers can focus on continually staying on top of network security, Yetman said.
Robert Gates covers data centers, data center strategies, server technologies, converged and hyper-converged infrastructure and open source operating systems for SearchDataCenter. Follow him on Twitter @RBGatesTT or Email him at email@example.com.
Best practices for secure data centers
Icelandic data center forgoes UPS, generators