GDCM Seeks to ‘Defrag’ the Data Center for Higher Efficiency
May 6, 2008 Alex Woodie
Earth Week has come and gone, hopefully leaving behind a desire to consume less and generally be more “green.” For operators of large data centers, cutting electricity consumption can not only gain you environmental kudos, but it can also save millions of dollars a year. That’s the goal of a little-known software company called GDCM, which is currently ramping up North American sales of nlyte, a Windows application that enables users to model their use of electricity, cooling, network capacity, and space, and predict when they will run out.
The GDCM story begins several years ago in England, where two IT guys for UBS, Lee Moreton and Robert Neave, sought better tools for managing the tremendous growth of the financial giant’s data centers. Finding nothing suitable on the market, Moreton and Neave decided to roll their own. What they came up with was the beginning of nlyte.
The two were struck by what they had created, so they soon left UBS to found their startup, GDCM–or Global Data Center Management–and began productizing their creation. After arranging financing and developing the first release of the product, the company started calling on some of the world’s largest operators of data centers. They were receptive, and today the company has several multi-million dollar contracts with some of the biggest financial services firms in the world under its belt, says Daniel Tautges, president of GDCM’s mid-market division, who was hired recently to help grow the fledgling North American business division.
Tautges, who is based in San Diego, California, says he has yet to run into a potential customer that doesn’t need a better way to manage its physical infrastructure. “That would be a new one,” he says. “CXOs are getting these power bills that went from $100,000 per year to $2 million per year. So these guys are freaking out. That’s probably an understatement.”
The problem can be largely summed up with one word: Blades. Sure, Hewlett-Packard and IBM will (correctly) tell you how much more efficient blades are, and it’s true that they do consume less electricity than their rack-based or floor-standing counterparts. Blades are very efficient with space, and have largely solved the space crunch in many data centers.
The real trouble with blades is keeping them cool. Because they’re so dense, it takes a lot more AC to keep them from melting. As a result, the fraction of electricity consumption attributable to cooling has skyrocketed. According to GDCM figures, it costs just as much, on average, to cool data centers today as it takes to power them. That is a new development with blades in just the last few years, and is fueling the exponential growth in data center utility bills.
Defragging the Data Center
So, what can GDCM do to reign in out-of-control spending on electricity to power and cool these racks of blades? In short, nlyte helps data centers optimize the layout of their data centers–defragging the data center, so to speak–to squeeze the most capacity out of their finite resources across four dimensions: electricity, cooling, network, and space.
“We have a visualization engine that looks at the whole data center framework,” Tautges says. “You can plan ‘what if’ scenarios as to moving equipment, installing equipment, and you can find where you have capacity very quickly, and also trend capacity over time. So for example if I continue to go up the same rate, when am I going to run out of power or cooling or space or network connections? If you’re out of power or cooling or space or network, you’re out of capacity.”
The software’s most useful capability may be assisting data center managers with devising strategies for boosting cooling efficiency. “It can tell you if you implement hot aisle, cold aisle, how much more efficiency you can get,” Tautges says. “Or if I go to virtualize these servers and eliminate these servers over here, what’s the best place in my data center to put the new server, based on what my capacity limitations are.”
Most data centers managers track their electricity consumption and cooling capacity using spreadsheets, but it can be difficult to bring it all together in a cohesive and intuitive way using that approach. “It’s not very advanced,” Tautges says. With front-end visualization tools connected to a back-end SQL Server database, nlyte brings it all together with an easy to use interface and advanced storage and workflow capabilities, the company says.
The nlyte software gathers data from many places, including a database that GDCM maintains containing the power and cooling requirements of thousands of servers. It also connects to existing CMDB (common management database) systems to pull other bits of information about server and network devices. Data defining other pieces of equipment can be pulled in through Web services interfaces. It also offers interfaces that unite the various parties typically involved in data center activities, including the application group, the network group, security, and facilities management.
The nlyte software can also provide insight into one seldom-considered characteristic of a data center: weight. With fully outfitted racks topping one thousand pounds, overstressing the design limitations of data center floors may be surprisingly easy.
Tautges recalls one customer that decided, almost as an afterthought, to check the floor stress levels before putting a new rack in. “It turns out before they dropped it in, they had to put a new joist in the data center on the 15th floor,” he says. The customer had to pay New York City contractor prices to install the joist, but it was a small price compared to the cost of a collapsed data center floor. Others haven’t been so lucky.
“Weight’s a big deal. It doesn’t seem like a big deal, but it’s a big deal,” Tautges says. “It’s all the physical characteristics of the DC. It’s an absolutely enormous problem for these folks.”
In the end analysis, nlyte can’t eliminate the core power requirements for running essential computers and network gear. A rack of blades that requires 500 watts is going to pull the juice it needs until it’s turned off. But nlyte can help data centers grow more intelligently, and position new equipment in the most efficient place in the data center.
Tautges says the typical nlyte customer gets a return on their investment (ROI) within 100 to 120 days. Merrill Lynch, for example, expects to make its money (about $3 million) back within 7 to 8 months, Tautges says. The bigger the datacenter and the more equipment churn it has (the average is 20 percent per month, Tautges says), the faster the customers will make their money back.
GDCM offers two versions of nlyte, including a mid market version designed for data centers with hundreds of servers, and an enterprise version that scales into the thousands of servers. Pricing for the enterprise version starts at $250 per server. For more information, visit www.gdcm.com.