|
 |
|
General Information
|
Add To My Personal Library |
October 31, 2008
• Vol.30 Issue 44
Page(s) 25 in print issue
|
Data Center Power Planning
How To Evaluate Your Power Needs
|
Jump to first occurrence of: [WEISMAN]
| Key Points
• Establish your equipment’s power and cooling profiles and understand your power densities on both a rack and a regional basis.
• Determine the cooling capacity available in your server room.
• Increase your power efficiencies through virtualization and energy-efficient equipment. | | Figuring out your organization’s power needs can be overwhelming. You can’t just plug devices into some surge protectors and wait for your utilities bill, just as you can no longer hook up some servers and storage into a closet and turn down the thermostat. Now you have to worry about your servers’ power and density profiles, target the direction of your cooling, and pinpoint ways you can save energy—often by using virtualization or other green measures—without compromising your network, among other things.
“It’s very common to run out of power or cooling, and in a lot of cases, customers are running out of both,” says Kris Domich, principal consultant for data center and storage solutions at IT services and solution provider Dimension Data (www.dimensiondata.com). “The amount of power being delivered to the room—or, more importantly, the amount of power that may be supported by a UPS subsystem—is often not tracked and so the capacity or available capacity may not be known or be readily available.”
So where do you start when trying to evaluate your IT power needs? Here are some suggestions to get you started.
 What’s Your Power Profile?
First off, get an understanding of what is driving your power needs today and in the future, says Domich. The best means to determine what these needs are is defining the power profile of your equipment. That way, as you add more hardware or swap out older servers for high-density servers, you will see the impact these actions will have on your power load and your cooling requirements.
Domich defines a power profile mainly in two parts: the wattage and heat density a given piece of hardware produces and then its overall cooling profile. “Does it draw air in from the front and expel air through the back, or does it draw it in from the rear or the bottom and expel it from another direction?” Domich says. “These are important things to consider when you place equipment into a rack because you don’t want to end up blowing hot air into the cool-air inlet of another system.”
According to Domich, different brands of servers, servers using different platforms, and servers of different ages will often have wildly divergent power profiles. A rack filled with standalone rackmount servers may use only a quarter of the watts that are needed by the same rack filled up with blade servers.
Handling The Load
Most data centers don’t drive things to a 100% load all the time; therefore, people mistakenly assume that their power distribution and cooling systems need to handle maximum loads. “In a lot of cases, you can get away with that, but you do need to understand when you will be driving exceptionally high loads, the durations you may be driving those loads, and whether your cooling and power distribution can accommodate those loads when, in fact, they are present,” says Domich.
To determine the maximum amount of power and heat removal that your equipment needs to run, Domich recommends checking out the manufacturers’ specifications. “These specs will show you the input voltage and the maximum amount of current that is expected to be drawn by that equipment. From that, you can make a calculation on the amount of power you will need, and from that you can make a calculation on the amount of heat, commonly measured in BTUs (British thermal units), that will be expelled by this piece of equipment,” Domich says.
Improving Efficiencies
It isn’t enough to figure out how much power you need. Joseph Thottuvelil, director of applications engineering and technical marketing at power conversion solutions provider Lineage Power (www.lineagepower.com), recommends purchasing high-efficiency hardware, as well.
When shopping for new equipment, Thottuvelil recommends comparing the power consumption of any equipment you plan to buy in addition to its computing and networking capabilities. According to Thottuvelil, this information should be readily available, but if you can’t find the information or have additional questions, contacting vendors directly is the most reasonable approach to take.
In addition to reduced power consumption, high-efficiency equipment tends to be more reliable. “Less heat is generated, leading to lower service and replacement costs,” points out Thottuvelil. Environmental benefits aside, there are clear economic advantages resulting from energy savings and the longer useful life of this equipment. 
by Robyn Weisman
Top Tips
Consider having an auxiliary power source. A generator can continue supplying your data center with power if you experience a commercial power loss, says Kris Domich, principal consultant for data center and storage solutions at IT services and solution provider Dimension Data (www.dimensiondata.com). In addition to the generator, consider using an uninterruptible power supply that will carry your data center’s power load during that short period (usually 30 seconds to a few minutes, depending on the generator) you are waiting for your generator to come online, Domich says.
Make sure your power delivery system can handle sudden spikes in power. After a commercial power loss, you need to take into account that your system will probably draw a significantly larger amount of power when it’s first turned back on, says Domich. Therefore, you want to make sure the power delivery system that’s supplying the electricity can accommodate that initial jump in electricity or introduce processes where you phase in or phase online groups of servers in groups at a time so that you don’t overburden your electrical system and blow a main breaker or breakers on your branch circuits because of that initial spike in electrical current.
Implement a targeted cooling strategy in addition to static pressure cooling. Domich says that a targeted cooling strategy, such as an in-row cooling solution or an enclosed rack solution, may be needed to remove excess heat away from the equipment and to be able to supply enough cold air to that equipment to keep it within manufacturer’s specifications for operating temperature. This point is particularly important if you use even a small percentage of blade servers in your network because of their power density. |
Reduce Power Needs By Going Virtual
As data centers become larger and the equipment more powerful, everyone is seeking out ways to use less power. One such strategy is implementing server virtualization, which plays a role in improving power and cooling efficiencies because you are consolidating the number of physical servers that you need, says Kris Domich, principal consultant for data center and storage solutions at Dimension Data (www.dimensiondata.com). As you virtualize, however, make sure you introduce levels of redundancy at the server platform level for an additional layer of protection against a server failure, given that you are relying on fewer physical servers to perform the same tasks, Domich adds. |
|
|
