|
 |
|
General Information
|
Add To My Personal Library |
June 19, 2009
• Vol.31 Issue 17
Page(s) 28 in print issue
|
Track Down Data Center Hot Spots
Monitor Temperatures To Protect Equipment & Reduce Cooling Bills
|
Some organizations waste large amounts of money pumping cold air into spaces inefficiently. Sometimes it is wrongly channeled and fails to arrive at the areas that really need it. In other instances, it is consumed by equipment lower in the rack so that upper servers have a tendency to overheat.
An important skill, therefore, is to be able to detect hot spots so you can take action to ensure they are adequately cared for.
“Concentrating proper airflow into the high heat load areas is critical,” says Jon Mills, a server monitoring specialist at Plixer (www.plixer.com). “If you know where the hot spots are, you can make sure the air conditioning is affecting the right areas.”
 Hands-on Monitoring
Hands-on monitoring is probably the cheapest yet most effective way to locate hot spots in a data center. Simply have an employee walk around the data center feeling for hot spots.
“The best, and easiest, way to detect data center hot spots is to simply walk around and feel for them with your hands,” says Vic Berger, a technologist at CDW (www.cdw.com). “The human body is pretty good at detecting changes in temperature, and in extreme cases, you will be able to detect areas that are hotter than others. Just walk around your data center, and you will feel them.”
This method can be supplemented with the use of a handheld digital thermometer. It provides exact readings and more accuracy in detecting temperature anomalies.
“We recommend placing the thermometer near the air intakes and fan vents of the devices to measure the changes in temperature,” Berger says.
Be Careful How You Measure
Like everything else, there is a right way and a wrong way to assess a data center for hot spots. Unfortunately, it is quite common for IT managers to take temperature readings in the wrong areas. One rookie mistake is to measure in the hot aisle and then add extra cooling in that area due to the presence of elevated temperatures. Therefore, you have to know where to measure and how to measure in order to detect actual hot spots.
“People read temperatures in the hot aisle and think that they have a hot spot or take readings in the wrong location in the cold aisle,” says Jeffery Evans, data center assessment project leader with the Liebert Services unit of Emerson Network Power (www.liebert.com). “You have to know the proper locations to take temperatures and then take daily readings in cold aisles.”
Taking readings at critical points is the first step in identifying hot spots and resolving problems that can result in equipment degradation. These measurements will determine whether heat is being successfully removed from sensitive and densely packed computer equipment such as blade servers.
“Start by taking temperature readings at the rack inlets in at least three locations per rack,” Evans says. “Begin low (6 inches off the floor), medium (3 feet off the floor), and then at a high level (top U-slot of the rack).”
Monitor Power Loads
It’s important to remember that power usage generates heat. Checking the power load of each piece of equipment, therefore, is a smart way to isolate possible areas of overload. Not only will an abnormal or elevated power draw indicate a possible hot spot, it also serves as an early warning for a struggling device.
“Checking the power loads of computer equipment is always a good idea,” Berger says, “although an important point to remember when analyzing power loads is that older systems tend to burn more power.”
Hardware vendors typically provide abundant documentation about the power load and recommended temperature ranges of their gear.
Deploy Temperature Monitoring Systems
Beyond a stroll down the aisle, taking manual measurements with a digital thermometer, or tracking power load, there are also more sophisticated temperature monitoring systems on the market that make use of a network of sensors. For SMEs, though, it might be best not to get too sophisticated. There are systems on the market that measure everything, slice and dice large numbers of metrics, and isolate the smallest anomalies. This might be needed in a large enterprise site, but simple is typically best for the average small to midsized enterprise.
Mills recommends a system with sensors that are easily relocated to enable more detailed testing and trending of suspect locations within the data center.
“Don’t go overboard with your temperature monitoring system,” Mills says. “Larger, more intrusive monitoring systems can take weeks or months to implement. A simple solution is often best because it offers more flexibility, and it is easy to set up and understand.” 
by Drew Robb
Bonus Tips
Use available tools. Tools are available inside intelligent PDUs and iLO (integrated lights-out) cards that allow you to manage the data center remotely and analyze how much energy it is using. Intelligent PDUs isolate areas of potential risk, as well as predict when a fan may be about to burn out before it actually does. This enables you to take appropriate preventive measures rather than only deal with reactive maintenance after something has gone wrong.
"Failing to use all of the tools at your disposal in power monitoring is an easily preventable error," says Vic Berger, a technologist at CDW (www.cdw.com).
Use computational fluid dynamics. Some SMEs are taking advantage of CFD (computational fluid dynamics), a simulation tool that graphically demonstrates the temperature characteristics across an entire computer room or data center.
“CFD provides a better understanding of why hot spots are present, as well as the effect of underfloor obstructions on airflow,” says Jeffery Evans, data center assessment project leader with the Liebert Services unit of Emerson Network Power (www.liebert.com). |
Best Tip: Do The Math
Don’t wait until hot spots appear, as this can result in damaged equipment. Be preemptive. Jon Mills, a server monitoring specialist at Plixer (www.plixer.com), lays out a simple 1-2-3:
1. Check each server’s heat output measurement against vendor documentation.
2. Determine a total heat load for each rack or area.
3. Ventilate and concentrate airflow into the racks and areas with the highest total heat load.
"The most effective way of locating hot spots is to simply take an assessment of the concentration of equipment and its associated heat load," Mills says. "IT hardware providers can usually provide heat load and cooling requirements for their hardware. It's a monotonous procedure but is really the best way to determine where the potentially dangerous areas are.” |
Best Return On Investment: What’s Hot & What’s Normal?
A common mistake is keeping devices at incorrect temperatures. Some think, for instance, that they have to keep every system at X temperature. A simple solution to this is checking the manufacturer’s specifications for each device. If you are trying to run it 10 degrees cooler than needed, you may be running up cooling bills.
“Reconsider your definition of hot,” says Vic Berger, a technologist at CDW (www.cdw.com). “Check the manufacturer’s specifications for each device, and you may be surprised to learn that many devices don’t need to be kept in a cold area. Super-cooling an entire data center can waste electricity and counteract any other cost savings you’re working to achieve.” |
|
|
