Page 1 of 2
The Data Center Dilema
In the tech world today, "going green" isn’t just a hats-off to Mother Nature. Sometimes it’s a sink-or-swim initiative on which a business' future lies -- especially when it comes to data centers.
Amid rising energy costs, data center operators are embarking on a quest for efficiency. The search is driven largely by financial interest, as electricity and cooling equipment imply hefty operational fees. For a lot of data centers, though, the need for efficiency is much more pressing. SMBs, especially, face physical space constraints and finite power supplies that -- if not maximized -- will stump their business' growth, eliminating any chance of expansion.
Data centers today consume almost 1.5 percent of the world's total electricity production and cost nearly $44.5 billion a year to run, according to industry analyst Linley Gwennap.
What’s more, DataCenterDynamics, another market researcher, projects that this consumption will rise a whopping 20 percent in 2012, putting total data center power consumption around 31 GW -- enough energy to power all the residential homes in France, Italy or the U.K.
If this growth estimate proves true, energy costs will skyrocket. "Looking at the growth projections for data center usage and the future of power generation growth, this trajectory is unsustainable," Gwennap warns. "A new paradigm for developing data centers based on energy efficiency will certainly help make data centers scale realistically with future demand growth."
Luckily, this new paradigm -- or at least a shift in its direction -- is here. Cognizant of data centers’ hunger for efficiency, solution providers in the processor and server space are taking significant strides toward the delivery of low-power, cost-saving alternatives. These advances, said to offer efficiency without sacrificing performance, will cut data center costs, provide a platform for SMB growth, and arm VARs with never-before-seen opportunities in the server and processor space.
Vendors Gear Up To Go Green
Intel, the market share leader in the processor space, hasn’t historically had the most robust offering of low-power processors. As data center consumption becomes a more prominent issue, however, the chipmaker has joined the ranks. In addition to its revamped Xeon 5600 processors, Intel has launched several server management solutions, intended to monitor and manage power and cooling resources within data centers.
One such solution, the Intel Intelligent Power Node Manager, brings component instrumentation to the platform level to optimize the use of every watt consumed. The second-generation release can report on system-level consumption, along with process and memory subsystem consumption.
Jay Kyathsandra, product marketing manager for data center and connected systems group at Intel, said that most data center racks peak at only 50 or 60 percent. Power Node Manager can bump those peak levels up to 80 or 90 percent.
"In a rack -- say you have a limit of five or seven kilowatts -- you can now go up to 80 or 90 percent of that full power capacity knowing fully well that, if there is a spike, the Power Node Manager will help the whole rack continue to run at a slightly dropped performance level. So you can peak up to the peak available power, and still not have an outage, which has always been a fear," Kyathsandra said. "We believe there is a lot of headroom there that can be capitalized with the right tools and technologies, while making sure that reliability and risk is addressed. That’s where Intel is coming from."
While Intel's power management solutions help data center operators control and optimize cooling procedures and performance, fellow chipmaker Nvidia is enabling energy conservation through the use of its GPUs and parallel processing architectures.
As GPUs continue to serve as low-power, high-performing complements to traditional CPUs within supercomputers, their ability to reduce costs within data centers is becoming more and more evident.
"GPUs are dramatically better when it comes to energy efficiency [compared to traditional CPUs]," said Sumit Gupta, director, high performance computing products at Nvidia. "And the number one thing I hear when I talk to data center guys is 'I’m power-limited. I’m limited in how much power I have in my data center, but the engineers, or the business unit, or the scientists are asking for more performance.' They’re feeling that pressure."
In response to this pressure, Nvidia has dedicated itself to the GPU, parallel computing model, and low-power ARM architecture, all said to offer more performance-per-watt than traditional CPU models.
And the chipmaker isn’t alone. PC giant Hewlett-Packard announced in early November a partnership with startup server vendor Calxeda in an effort to bring less power-hungry ARM-based servers to data centers. The initiative, now deemed Project Moonshot, has thrust HP to an influential, front-and-center position within the move toward efficiency.
As part of Project Moonshot, HP unveiled its Redstone Server Development Platform based on quad-core EnergyCore ARM Cortex processors. The development platform will serve as testing grounds for customers and partners to explore new avenues for data center power reduction.
NEXT: AMD Steps Up
