Have we really kept pace with what should be required or what works best?

By Peter Hildebrandt

We live in a world of change when it comes to computers and data centers. The chance that today’s great new idea will be standard equipment on computer technology in six months or less is a growing reality. But what about the thermal needs and demands of such equipment?

Airflow issues in information technology, computer, or medical treatment environments may be low on the priority list. After all, if everything seems to be running fine—with all results normal and equipment not failing—what more could be asked for? Evaluating a center’s airflow sounds extremely complex. Is there a problem? If you’re not sure, why bother?

Well, it turns out computers do have an optimal temperature for operating, and the HVAC at a data center can be designed to enable a computer system to best meet that temperature through its airflow—despite how complicated a system might be. Similar in some ways to an uninterruptible power supply’s (UPS’s) ability to help a system stay up and running, proper thermal transfer throughout an operation helps a computer system run at the level for which it was designed, ultimately prolonging the computer’s lifespan.

Also, its airflow applications can be modeled with software—3-D animation software, specifically—to understand the various factors influencing airflow. This can be done systematically to identify the many interrelated aspects of a particular pattern of airflow.

Again, though: Why bother? Well, one good reason is that such software is able to optimize design estimates of air volume and set point temperatures to enable impressive savings in capital investment and system operating costs. Another is the software’s capability for predicting and eliminating airflow problems or hazards early in the design process, before building or renovation begins.

It’s also possible to successfully troubleshoot airflow problems or hazards in and around existing facilities to ensure fast resolution with minimum downtime.

Working RCI Onto Radar Screens
Hoping to accomplish some of the things mentioned above, one major telecom company turned to ANCIS Inc., a consulting firm based in San Francisco. Magnus K. Herrlin, president and founder of ANCIS, has 10 years of experience with the research-and-development (R&D) efforts of the regional telephone companies, where he saw the need for advanced services for both the data center and telecom industries. “We want to understand how these facilities should operate,” says Herrlin.

The rack cooling index (RCI) addresses the thermal aspect. (“Rack,” in the computer industry, refers to the cabinet or place where the computer equipment is located.) The RCI is a measure of how well you have managed to operate your environment within a particular range, say, 68°F–77°F. If all equipment senses a temperature within that range, your RCI is 100%. “But if some computers experience a temperature too low or too high outside that range, then the RCI is going to drop to 50% or sometimes even as low as 30%,” Herrlin says. “The RCI is in effect telling an owner how well they have managed to maintain their equipment. This will probably be reflected in the equipment’s reliability and uptime as well.

“There seemed to be a need for services for the data center industry and the telecom industry alike. And this is the kind of work we are doing today; we do a lot of writing of standards for different companies on how to manage their data center or telecom environments.”

Herrlin was hired into the telecom industry’s R&D arm in 1993 to take care of two things: energy issues and thermal issues for the seven regional phone companies. Despite this fairly high workload for Herrlin, the experience also resulted in more than 30 specifications for the phone companies, including how to design, build, and maintain energy in thermal systems.

The RCI describes how well the servers in a computer system are cooled. This is important because the reliability of computer equipment is dependent on efficient cooling, though it doesn’t have anything to do with the energy consumption of the data center as a whole. An RCI of 100%, for instance, does not necessarily mean that energy consumption has been optimized. It could actually be the opposite, according to Herrlin, with extra power being used to cool the computer system. “The RCI is simply describing how well the equipment is cooled,” he says. “A computer system is much like a human being in one respect. As with us, if our environment goes too much over 75 degrees, we start to feel uncomfortable.

“On the other hand, if it’s lower than 60 degrees, that will feel cool. Likewise, a computer has its own limits: With too much heat there’s the potential for certain internal failures, and, if it’s too cool, other types of failures and more energy will be spent to get the system back up to the optimal temperature.

“In effect, the chiller system has to work harder, and this will cost you more. Systems often do not have specs for lower temperatures. If such a range is reached, something termed ‘timing problems’ may occur; the computer is not going to calculate in the right way.” The industry does have guidelines as to what the optimal temperature is for operating computer systems.

State-of-the-Art Technology in “Model T” Cabinets
The computer systems that ANCIS is cooling are usually installed in structures called “cabinets.” These often have a 2-foot-by-2-foot footprint and a height of 6 feet. The cabinets also tend to be extremely hot. For certain types of computer equipment, one single cabinet can contain 30,000 watts. Although not on the market yet, cabinets containing up to 50,000 watts are in the pipeline. Most computer or data center environments were not built for such heat loads.

“This is why this work is very important,” says Herrlin. “There are no real good solid guidelines for companies to follow when it comes to these heat loads. The various computer technologies involved materialize only to be light-years ahead of the physical plan of the facility or basic cabinetry for containing the computer equipment—sometimes still from out of the 1970s.

“Add on to that dilemma the fact that no one knows what is coming on in a few more years from the manufacturers. There is a real disconnect between equipment manufacturers and the service providers who own and operate equipment.”

Through the Energy Protection Agency, the federal government has recognized that modern computer systems are consuming great amounts of energy in an effort to run efficiently. The equivalent of fuel-efficiency standards for cars has not yet been established for computer systems, according to Herrlin, so no one knows how efficient a computer is. Currently there is a new initiative at the EPA to develop an energy-efficiency standard or specifications for computer systems along a path similar to that of the standards already developed for automobiles.

“If you get a spec for an end user, a buyer of a large computer system, and this individual is choosing between two systems whose performance is absolutely the same with the exception that one consumes half the energy of the other, the purchaser might want to consider buying the latter system. Energy costs in a data center are especially big portions of the overall budget.”

Almost every major organization in the United States and around the world has a small or large data center. They all have the same problems. It doesn’t mean you have thermal problems just because your data center is large; you can have the same problems if you have a small data center.

Software Application Company Assists
Flomerics Inc., a design software firm founded in 1988 with headquarters in Marlborough, MA, teamed up with ANCIS to solve the thermal flow and HVAC problems for the telecom company. Flomerics used FLOVENT software, which is geared to facilities and architectural applications, for this major telecom project.

Commercial products for computational fluid dynamics (CFD) have been around for over 30 years. When a constructed environment such as a room is studied, there are two ways to look at its transfer of air and energy. One is to look at it as a whole—as a “black box” with equations applied in the study of energy consumption or air quality.

The other way to compute CFD for that same room is to divide it up into thousands or millions of segments or volumes, commonly called cells. Then, instead of trying to calculate the whole room as one single entity, the calculation is done simply from one cell to the next. Calculation is done for the heat flow from one temperature to the next and also for the airflow in an effort to determine temperature and airflow in each of these cells.

When the calculations are taking place, such as for a row of cells, this continues over and over throughout the room, making it an iterative process. Recalculations are done until equilibrium is reached. This then tells the investigator everything that is going on in the room.

The original equations for CFD were developed in the early 20th century, according to Christopher Wark, FLOVENT technical sales manager. “The mathematics involved was developed along with that of computers; therefore, the computer program algorithms have also been improved along with the computing power,” says Wark. “Being able to predict what that temperature right at the equipment intake is, along with everything else in the room, becomes far more efficient and more accurate as time goes on.”

As CFD computer programs evolved, they were primarily used in research and, from a commercial standpoint, for aerospace, engine design, and some other automotive applications. “But the programs required some serious expertise to be used,” says Wark. “They were not easy to use, and some remain so today.

“Flomerics developers believed CFD could be simplified tremendously. The idea was to get CFD technology out of the research and development lab and get it into the hands of all engineering designers.”

Through the FLOVENT software product, Flomerics has made CFD easy to use and highlighted its use in settings for which it was particularly well suited, such as data centers or other facilities.

Cooling and HVAC Operations
One of Flomerics’ biggest customers is American Power Conversion, a company known for its UPS products. Because telecoms must be kept up and running, they tend to have huge UPS operations.

“The issue arising may not be simply, ‘Does your UPS keep your power for your electronics going?’ but instead, ‘Is your UPS big enough to keep your cooling going?’” says Wark. “One of the important features of FLOVENT is its ability to do a transient analysis: Starting with an equilibrium condition, something changes and you study what happens over time until you hit the next equilibrium condition. In other words, everything’s up and running, everything’s full-speed ahead with all your equipment cranking away—and then the power goes out. You have a UPS system keeping the computers alive, but you also just lost all your air conditioning.

“How long do you now have before you must either shut things down in an orderly fashion or get your air conditioners up and running? Our FLOVENT software tool enables users to see exactly what’s going to happen in such a case.” It’s used not just for optimizing the airflow when things are normally running, but it can also be used to see where things get hot when things start to shut down.

One of Flomerics’ customers is currently upgrading its computer equipment. “But they upgrade to a certain extent until they’re not exactly sure of where they stand with their facilities,” Wark says. “They may have some questions about whether they are still confident everything is working the way they intended to.

“In the case of their super computers, when they start them up, it’s such a tremendous drain the local power company knows immediately when someone onsite is doing a calculation. The customer is going to want to know that everything is coming up together and there isn’t the lag with their air handling as the computers heat up.”

Other cases include remote telecom repeater stations in the desert, where, despite excellent insulation, the sun must be taken into consideration as a factor with heat load in addition to the cooling of the site’s generator. FLOVENT software can even address these situations, as well as the problems involved in designing the facility and the HVAC system.

“When ANCIS worked with the large telecom, Dr. Herrlin looked for the ideal temperature; there were also all these other things going on which will affect that temperature,” says Wark. “Here is where the FLOVENT software comes in. Some of these issues include that of more current moving through a particular area of concern. Others are how well the cooling air is reaching the intake for the computer in the face of beneath-the-floor obstructions, floor tile design, or how well the air-conditioning intakes themselves are located.”

The FLOVENT software is a very powerful product but very easy to use, according to Wark. The idea is that any engineer with a rudimentary understanding of what heat transfer is all about should be able to use it. It is designed to run on PCs as well. FLOVENT’s current market is largely engineering firms doing data center design.

The software contains an optimization tool for dealing with the computer room air-conditioner rack (CRACK). The best location for that CRACK unit can be found through a design experiment feature included in the FLOVENT product. Once the location is entered, the optimum temperature range is also added before clicking “Go.” The software then needs to be left alone for a time in order for it to go through a number of iterations. After this, it will have located the optimum spot to place the CRACK unit.

When it comes to modeling air handling, the air must be coming through the floor properly, the CRACK units must be in the right locations, and the flow rates of those units must be high enough.

Getting Rid of Frigid Computer Centers
If air can be moved from the source to the piece of equipment as quickly and in as short a path as possible, then the overall temperature can be raised and you don’t have to have the typical “refrigerator” computer center. Workers are more comfortable, and energy is saved.

“The fact that you are working in this open space is why you need CFD to accomplish this,” explains Wark. “For computer rooms and data rooms generally, floor obstructions have been noted as a big problem.

“Huge masses of cables shoved beneath the floor have become a huge issue. Cabling must be bundled sufficiently and not obstruct airflow. There are now companies simply focused on having a tidy, obstruction-free data center, because all of it affects the efficiency of what you’re doing.”

Apart from data centers, the most common applications for the FLOVENT software are in clean-room designs—such as those found in ICUs or even in the National Institutes of Health—requiring air curtains. These can be patient rooms, operating rooms, or laboratory environments. Those applications are in addition to auditoriums, schools, office spaces, and a broad range of settings.

Because of all its potential for energy savings, another use with a great deal of current attention is that of natural ventilation. “In summary, this simply comes down to opening a window in a room—but also in a logical manner, from an energy standpoint,” says Wark, who is also involved with the US Green Building Council. “Is the thermostat going to be coming on at the wrong time? It has now been found that the only way to be able to model a building for optimized natural ventilation is by using CFD.

“A company might be strapped for money, and therefore they also spend much time putting out fires instead of planning ahead for how money can be saved two, three, or five years down the road. But there are organizations which think about these things.”

Herrlin gives high marks to some US companies that have worked hard on their own in this area. Some of the chip manufacturers, for example, have for many years worked to develop more energy-efficient chips.

Throughout 2006, Herrlin gave numerous presentations at different conferences.

“At all these different meetings they were discussing the issues of energy, power, and thermal management in computer and telecom facilities,” he says. “I believe these will continue to be the salient issues for the industry for some years to come, so we will continue on the path we are on and provide cutting-edge advice for companies.”

DE - May/June 2007