Demands Of The New Data Center: Consolidate, Virtualize, Cool4:28 PM EST Fri. Mar. 18, 2011
Energy cost savings were increased from to 54 percent from 27 percent by targeting UPS and cooling systems. Click image for full-sized view.
During tests conducted recently by the CRN Test Center of Dell's C6145 PowerEdge server, which packs two server nodes, 96 processor cores and 1 TB of memory into a single 2U cabinet, we began to think about how such server density might compare with cutbacks in a public transportation system. As more and more people cram into fewer and fewer buses and trains, each vehicle must consume more energy to bring riders from Point A to Point B.
Aside from the obvious issues -- longer lines, fewer seats and other innocuous problems that are borne by hapless riders, increased energy consumption and its associated costs are often overlooked by the central planners, who argue over how best to spend new-found cost savings gained from cutting public transportation by half.
The same thing can happen in the data center in a number of ways, and can occur even if you're paying close attention to the efficiency and power usage of devices that you're deploying. "Typically, data centers were built years ago and were not designed for today's efficiencies," says Philip Fischer, data center business development manager at APC by Schneider Electric. Simply adding modern, efficient equipment to an existing data center can reduce or limit overall efficiency if it's not managed properly.
Fischer described a case study in which one company consolidated and virtualized the servers in its data center by a ratio of about 20 to 1, which alone reduced power consumption cost by about 27 percent (see Figure 1). "It's interesting to note, however, that their PUE went from 2.04 to 2.56," said Fischer, referring to the power usage efficiency figure that results from dividing a facility's total power usage by the energy used by just the IT equipment. Lower PUE numbers indicate higher efficiency.
"Their PUE declined because they had a higher ratio of wasted energy to the actual energy used by the IT equipment," he said. After modernizing some infrastructure equipment and right-sizing some others, the company brought its PUE down to 1.61. Google is considered an industry leader in data center energy efficiency with a PUE of 1.21. "After they went one step further to optimize the power and cooling with close-coupled systems, they were able to increase savings and efficiency much more."
The measurements were performed using APC's TradeOff Tools (tools.apc.com), a set of free calculators that the company offers resellers and end-users for figuring out power usage, UPS requirements, carbon footprint size, and myriad other power-related functions.
Spending on servers, power and cooling has risen steadily since the mid-1990s, with emphasis on power and cooling, said IDC. Click image for full-sized view.
APC also offers fee-based tools that resellers might use to build a managed service practice. InfraStruxure Central is a CRN Test Center-recommended product that provides a graphical representation of a data center, ready to populate or modify with the customer’s server, storage, power and cooling assets. Capabilities of the centralized management tool include simplified planning of power distribution and hot- and cold-aisle isolation setup and management, monitoring and reporting of power consumption and efficiency and implementation of what-if scenarios. An alerting engine permits warnings to be issues if levels reach critical thresholds.
Fischer spoke of the huge opportunities that exist for resellers with just a little specialized knowledge to answer the call of customers looking to right-size their data center. "We've seen many customers surprised after virtualizing that did not take data center power usage into the account." Resellers can use APC's TradeOff Tools, he said, to measure the effects of virtualization on a customer's existing infrastructure and formulate a strategy for growth and efficiency.
On the front lines of integration with virtualization is Eaton, whose impressive tools for power management are more tightly integrated with VMware than anything we've seen. Also recommended by the CRN Test Center is the company's Intelligent Power Software Suite, which consists of Intelligent Power Manager and Intelligent Power Protector modules. IPM monitors and manages Windows PCs and servers as well as systems running VMware vCenter and Microsoft SCVMM. IPP takes care of shutting down systems following extended power outages and works with certain versions of Linux as well as VMware ESX and ESXi and Windows Hyper-V.
IPM can be integrated into vCenter and "will kick off vMotion, so if there's a power disturbance in the network, our software will kick off that software for you and ensure that there's no interruption" by automatically migrating any running virtual machines to a machine that's not in distress, said Curtiz Gangi, director of U.S. channel sales for Eaton. "We're also logging those events through vCenter on the server dashboard. We're the only provider that's actually part of the vSphere platform, and you can see our data right there in one viewing screen. It also permits graceful shutdown of ESX and ESXi and its virtual machines."
We tested Eaton's VMware plug-in, and liked the convenient visibility it offered administrators, who have the ability to observe and manage power devices right through the vCenter Server console. Of course, the software also sends alerts and notifications of power events at user-specified levels. Other benefits of power management are to maximize and enhance system availability, simplified asset management, remote on/off control of connected devices and the ability to perform power analysis and trending reports. "Those are key to helping your customers manage their virtualized environments" or for providing those services for a fee.
Gangi also spoke of the power consumption as an afterthought, saying that it's among the most overlooked characteristics of increased data center density. "The hidden risks related to power can be overloaded circuits, breakers tripping and a rise in power anomalies because we're drawing too much from the power environment. We're seeing real issues in deployment." Redundancy also is key to high availability. "Not having a single point of failure on any critical infrastructure equipment is paramount to the success of those networks and their deployment of data center and distributed networks."
According to IDC (Figure 2), world spending on servers, power and cooling is on the rise, and that trend has continued almost unabated since the mid-1990s. According to APC's Fischer, an increasing percentage of that spending is going into power and cooling. "They're not just throwing money at the problem, they're looking for a systematic approach to managing data center cooling and power usage," he said. This presents reseller opportunities for only for revamping a company's entire data center, but also to take steps toward consolidation, increased energy efficiency, higher capacity and better management as companies seek to decrease the resources necessary for infrastructure management.
Data centers are not limited only to the Fortune 500. "A common misnomer about data center growth is that they're always large," said Fischer, who says that upwards of 80 percent exist in the SMB. "This represents a huge opportunity for helping with changes, adding capacity segmenting the data center," and for the implementation of virtualization and other new technologies. "Customers are facing problems they never had before, and they're not experts. So they're looking for partners."
As data centers continue to consolidate, the need for increased power and cooling will become apparent, either by accident or by design. Resellers stand to benefit from this trend either way.