The First Facility Management Blog

This Web Exclusive article has been contributed by Carl Cottuli, vice president of products and services at Wright Line, an American manufacturer of air flow management systems based in Worcester, MA.

Data Center Air Flow: Measure It, Then Manage It

Per square foot, a data center within a larger facility is the most expensive portion of a modern building complex. This expense is due to evolving data center energy costs. Often it is the facilities manager who is asked to manage the data center’s energy needs.

Many data center managers understand that cooling improvement is always possible, but it is never clear to know how much improvement is possible without having to tear down and build new. The following is a short, simple, yet definitive, method of quantifying the energy/air flow status of your data center.

To improve cooling, you’ll need to identify four critical temperatures.

• Temperature 1 — A/C Outlet: Gather and average all air conditioning outlet temperatures at a given time.
• Temperature 2 — IT Inlet: Gather and average the lowest, mid-level, and uppermost rack inlet temperatures for a good statistical sampling for the rows in the data center.
• Temperature 3 — A/C Inlet: Gather and average the warm air stream returning for all air conditioning units in operation.
• Temperature 4 — IT Outlet: Gather and average the lowest, mid-level, and uppermost rack outlet temperatures for a good statistical sampling  for the rows in the data center.

Subtract the Temperature 1 from Temperature 2. The result or “delta” is the average temperature rise in the supply chain of cool air entering the data center. This indicates how much heat is entering into the cool air stream before it gets to the IT equipment. Controlling this type of waste is important and will drive an improved performance for cooling the IT equipment.

If your Supply Path delta temperature is high, such as 15 degrees, for example, this represents a potential amount of energy to save and use elsewhere—such as for new IT equipment roll-outs—after an air flow management strategy is employed to eliminate some or all of the waste.

Next, subtract Temperature 4 from Temperature 3. This number will show the return path delta temperature, and will indicate the amount of cool air that is bypassing the IT equipment and returning directly to the air conditioning unit without cooling any IT equipment.

A significant change in the return path temperature above a few degrees indicates that some amount of chaos cooling is present in the data center and can lead to significant unpredictability in cooling performance.

These quantitative readings should be evaluated as pairs of delta temperatures in the supply and return path. A data center with many times more cooling capacity than demand would yield a very low delta in the supply path, but a very high delta in the return path. Consequently, there would be an opportunity to improve air flow management in the system. Also, when making any improvements, it is always very important to be mindful of the impact any air flow changes have on both supply and return optimization.

Cottuli can be reached by e-mail at carl.cottuli@wrightline.com.

Practice has begun, marking the beginning of the 2009 FIA Formula One World Championship. The first race of the season is being held this Sunday, March 29 in Melbourne, Australia in Albert Park.

The course there, as described on the Formula One Web site, is a “technically demanding, high-speed ‘street’ circuit.” Cars reach maximum speeds of 300 km/h (185 mph) and average 225 km/h (140 mph) around the 5.3 kilometre (3.3 mile) 16-turn lap. The track surface is bumpy, particularly in the braking areas, hard on the brakes and slippery at the start of the race weekend before the road rubbers in.

Twenty drivers from these 10 teams practiced on Friday (in Australia), with their standings as follows after the second round of the day: #1 McLaren-Mercedes; #2 Ferrari; #3 BMW Sauber; #4 Renault; #5 Toyota; #6 STR-Ferrari; #7 RBR-Renault; #8 Williams-Toyota; #9 Force India-Mercedes; #10 Brawn-Mercedes

The Ferrari Action
In a recent trip to Maranello, Italy, I had the opportunity to visit the data center for Scuderia Ferrari, the division of the Ferrari automobile company concerned with racing (the Ferrari team finished the 2008 Formula One season in 1st position). The data center, located at the Ferrari headquarters, was put into operation in 2005 and supports the IT operations of the company’s Formula One activities.

This steering wheel from a 2006 race car shows the numerous controls used by drivers.

The cars that are raced in Formula One events are high tech machines. At each race, Scuderia Ferrari sets up an on-site data center where team technicians track their car’s performance during a race in real time. They can then communicate information to the driver, enabling the driver to track all aspects of the car’s performance (i.e. acceleration, braking) as they compete. Gathering this information also enables the team to improve performance for upcoming races.

The data center at Ferrari plays a central part in this exchange of information, and as Piergiorgio Grossi, CIO, Ferrari explained to the group during my visit, the data center needed to provide two things—flexibility and reliability (he expected more than 99.999%).

The facility was built to house both medium- and high-density operations (drawing up to 20kW per rack); to that end the designers took a “box within a box” approach, housing the high-density, more cooling-intensive, section separately within the larger data center space.

Flexibility in configuration, capacity, and cooling led Ferrari to utilize APC by Schneider Electric power and cooling equipment (APC also oversaw the construction, except for the IT equipment). The company chose to use APC’s Row Based cooling and built the facility using the APC InfraStruXure standardized methodology.

Piergiorgio Grossi, CIO, Ferrari

During the tour, Rob Bunger, director, business
development enterprise & systems, North America at APC, noted, “Ferrari was an early adopter of Row Based cooling and built a data center using the APC InfraStruXure standardized methodology. Not only have they reliably powered and cooled high density loads, but they have also been able to grow with IT refreshes and improvements to the power & cooling with minimal disruption to their operations.” Commenting on the well maintained appearance of the facility, Bunger pointed to Ferrari’s implementation of standardized solutions, along with their operation procedures.

Scuderia Ferrarri data center in Maranello, Italy

Massimo MartelliRossi, data center manager at Ferrari, has overseen the space for the past four years, and he takes great care to ensure the facility is in in tip-top shape. Several times during our visit, he used a trusty device to spot check the temperature and humidity within the rows of server racks (he can also check this data from his desk).

Cabling is tidily bundled above the racks. MartelliRossi explains this makes for easier access to change out components in the center. He also explains that Ferrari’s decision to locate cabling above the racks, rather than under floor, avoids two potential problems—damage to cable, in case the two large cooling pipes below ever burst, and prevention of rabbits and other creatures from chewing on them.

The data center of Scuderia Ferrari is in the thick of its high season. For the next eight months, throughout the Formula One season, this high tech gem in the Italian countryside will keep things moving on racetracks throughout the world.

According to a survey released earlier this week by Siemens, many companies are uncertain about how to identify opportunities and reach goals for energy efficiency, especially in the realm of their data processing and storage. The study, which examined general IT efficiency attitudes and practices, revealed that nearly three quarters of the Fortune 2000 respondents (87%) believe it is important to pursue overall energy efficient practices, but only 48% have a stated goal to reduce their carbon footprint, and even less have begun to take action.

“As businesses continue to search for ways to save money and alleviate the strain on the power grid, green technology across the board is becoming a necessity rather than just ‘the right thing to do’—examining and implementing data center efficiencies is one major area we cannot overlook,” stated Ken Cornelius, CEO of Siemens One, a business unit of Siemens AG. “If we do not start looking closely at our data centers now, 70% of the world’s data centers will have tangible system disruptions by 2011 and the systems will experience world-wide brown outs over the course of the next five years.”

Data centers account for 2.5% of the world’s energy use, which is expected to grow by 12% a year, placing more strain on the power grid. The EPA notes that the energy used by data centers in the U.S. is more than the electricity consumed by all of the nation’s color TV sets and comparable to the electricity consumption of approximately 5.8 million homes. According to the Siemens survey, a majority (65%) of leading Fortune 2000 companies recently reported that the costs of running their data centers have increased over the last few years.

Seventy-two percent of those surveyed said the size of the investment was the primary barrier to improving their data center energy efficiency, followed by lack of specific information on the return of the investment for making changes (38%), server down time required to implement changes (38%), and concerns about running legacy software on new systems (34%).

Respondents also showed concern about their data centers exceeding processing capacity and the risk of losing their data. About one-quarter (27%) worry a “brown out” or exceeding process capacity will affect their data centers, while 27% are worried about losing data, 15% worry about aging facilities, and 10% expressed concern about high energy costs.

“The tools and technologies are available to help companies get started—whether it is simply through server virtualization, installing power monitoring systems, retrofitting an existing structure, or with new construction,” added Cornelius.

The above findings were the result of a survey commissioned by Siemens Corporation and KRC Research from October 23 to November 24, 2008. A nationally represented sample included top, high, and mid-level executives of Fortune 2000 companies that are in some way involved in making decisions about their companies’ data centers. Because Fortune magazine does not publish a list of the top 2000 companies, the sample was created by extending the Fortune criteria: the 2000 largest American companies, ranked by annual revenue.

(Photo courtesy of Siemens)

The tech sector has not been immune to the recession, although one area—data centers—continues to evolve at an impressive pace. Once known for its “down and dirty” personal computers, Dell is now applying a low tech solution to the data center concept in order to gain some market share in this highly competitive arena.

Dubbed “Humidor,” the custom containerized data center was launched by Dell’s Data Center Solutions unit to accommodate the intense demands of Microsoft’s Azure compute cloud. While the structures are capable of meeting this need, they also enable Dell to avoid the construction of thousands of square feet of permanent brick and mortar data centers.

Cross section view of the Dell "Humidor" (photo from The Register); bottom racks support any IT gear. Flexibility in this section could translate into a 10-year "shelf life" for the container.

Timothy Prickett Morgan of the Register (12/13/08) writes:

Andy Rhodes, director of marketing for Dell’s DCS unit explains that hyperscale customers are interested in compute and storage density, and they don’t want to spend a lot on data center facilities….Moreover, the facilities people and the IT people do not always get along, causing HR friction in data center solutions where power, cooling, and IT are mixing inside a single container. Each group needs different—and separate—access to their respective gear.

The top part of the Humidor double-container has power transformation, metering, and distribution gear, uninterruptible power supplies, and air handlers (FM-related equipment), according to Rhodes. And the facilities people apparently don’t mind having to take the stairs to their own gear so long as they don’t have to share with the IT nerds.

Dell expects customers to put a cheap shell of some sort around the containers as they are lined up and stacked, not to leave them in the parking lot. Having millions of dollars in IT gear sitting exposed, without physical security and inside portable containers, is not the brightest thing anyone has ever suggested. But using containers to cut data center facility costs by anywhere from 20% to 30% is pretty smart, which is what Dell reckons a containerized data center can yield over a conventional brick-and-mortar data center, depending on a lot of variables, of course.

So the next time you drive past a parking lot or storage area filled with innocuous looking containers, you may be inclined to do a double take. After all, they may not be stuffed with cheap commodities from overseas—they could be fully functional and highly sophisticated data centers, crunching your numbers and chugging away as you drive past them on the highway.

Emerson Network Power, a business of Emerson, released its annual Data Center Users Group (DCUG) “Inside the Data Center” report yesterday. More than 165 data center, facility, and IT managers participated in the survey, which reported these findings:

• Data center criticality is rising. Three out of four respondents said in this year’s survey that the data center is more critical today than in past years. Many of these are large facilities that have had high availability goals for years.
• Power densities across the room and within the rack are increasing sharply. In doing so, they’re pushing facility capacity to the max—and fast. In the last two years, average rack power density has risen from 6 kW to 8 kW. More than 10% of respondents said they will be out of data center capacity by the end of this year, and a total of 68% percent expect to be at capacity within the next three years.
• Despite tight capital equipment budgets, changes to the data center are still planned. Thirty-seven percent of survey participants responded that current economic conditions are impacting their ability to operate or expand their data center. However, in a sign of just how dynamic and critical the data center has become, more than 75% of survey respondents are still planning for changes to take place within their facilities.
• Data center managers are unwilling to compromise availability for efficiency. Data center managers will be taking into account corporate initiatives to improve energy usage throughout the data center; yet, this directive is challenged by the understanding that availability remains the top priority. On a continuum of one (reducing energy usage is main priority) to seven (maintaining high availability is our main priority), 69% of respondents identified a priority level between five and seven in favor of high availability.
• Energy usage strategy still missing. A comprehensive energy usage strategy remains elusive for a majority of businesses despite one in four respondents having completed an analysis of the efficiency of their data center equipment. Only 28% of survey respondents have a documented strategy to reduce energy usage. 

The entire DCUG “Inside the Data Center” report is available for download in the white papers section of www.liebert.com.
 
The DCUG is a group of approximately 2,000 data center, IT, and facility managers founded by Emerson Network Power in 2003. The group meets semi-annually to collaboratively discuss the most relevant issues affecting the reliability, availability, and cost of operation for mission critical installations. The group’s membership comprises executives with a wide variety of IT and facilities management expertise from an assortment of companies, including board member companies Vanguard, Cincinnati Bell Technology Solutions, and JPMorgan Chase, among others.

Given the current economic conditions, here are 10 suggestions for facility managers to get more from their data centers in 2009 while spending less money, courtesy of Emerson Network Power.

1. Cover Your Bases. It may be more difficult to recover from an outage during tough economic times than during prosperous one. A relatively small investment in precision air conditioning and backup power can actually save money. For example, precision air conditioning will adequately protect data center assets; building air conditioning alone will not. A double conversion backup power solution with adequate redundancy is essential to raising system availability and ensuring business continuity.
2. Look Inside Before Outside. Increasing density may be a more cost effective approach to meet the need for more capacity than new facility development. For example new cooling architectures can enable densities notably higher than average data center densities at a fraction of the cost of building a new facility.
3. Assess Before Action. Perhaps one of the smartest investments businesses can make in the coming year will be to asses their data center to identify and resolve vulnerabilities that threaten availability, increase data center efficiency, and improve planning and budget allocation.
4. Go From Room to Rack. Utilizing an integrated enclosure system (i.e. data center in a box or mini computer room offers a cost effective solution to protecting the equipment that may be in a small data center or room. Instead of conditioning the whole room environment, just protect the rack.
5. Cap the Cold Aisle. Cold aisle containment allows cooling units to run at reduced capacity to achieve ideal cooling conditions and save energy costs. This tactic is more efficient and effective than hot aisle containment systems and offers a better environment for data center personnel.
6. Check the Weather Forecast. In many locations, economizers can be used to allow outside cool air to complement data center cooling systems and provide “free cooling” during colder months. This approach lowers energy usage, lessens wear on some components in the cooling equipment, and decreases operational costs. All together, it can be a welcome reduction in the data center electricity bill.
7. Watch Often—If Not Always. The importance of monitoring what’s going on inside the complex and dynamic data center is more important than ever. Keeping an eye on performance will help businesses steer clear of unnecessary maintenance and repair costs. Success in this endeavor will require IT and facilities to integrate disparate data into a centralized portal where actionable and meaningful information can be derived.
8. Improve Energy Utilization. Opportunities exist to improve energy use throughout data centers of all sizes. For example, adding variable frequency drives to cooling systems allows them to recognize reduced loads and operate more efficiently. Every watt of savings achieved on the processor level will create a total of 2.84 watts of savings for the facility.
9. Avoid Cutting Corners. A preventive maintenance plan can extend equipment life and reduce maintenance costs. For example, employ a battery maintenance strategy so that your business isn’t a victim of the number one cause of UPS failure: bad batteries.
10. Don’t Stop Thinking About Tomorrow. It may be necessary to minimize capital expenditures but make sure you don’t compromise future scalability. UPS scalability is emerging as a popular solution to reducing the risk associated with miscalculating future capacities. Statistical analysis of UPS system configurations in light of failure rates shows that system reliability begins to decrease sharply when more than four UPS modules are used in a single system.

As part of its work, the California Energy Commission operates a Research Development and Demonstration (RD&D) Division, which oversees the Public Interest Energy Research (PIER) Program. The PIER program focuses on a number of energy areas, including Building Efficiency. This research aims to decrease building energy use by developing or improving energy-efficient technologies, strategies, tools, and building performance evaluation methods.

Periodically, PIER will publish informational briefs based on its research, which is often conducted with a variety of organizations with funding from the California Energy Commission’s Public Interest Energy Research (PIER) Program. The briefs are written and prepared by E Source.

PIER recently released a brief on research on the efficiencies of various types of uninterruptible power supplies (UPSs) for data centers under a variety of operating conditions. The study, titled “Uninterruptible Power Supplies: A Data Center Efficiency Opportunity,” also proposed an efficiency label for UPSs and estimated the current energy use of the existing stock of UPSs and the potential savings if efficiency standards were in place.

Collaborators on the project included Lawrence Berkeley National Laboratory and its subcontractors, EPRI Solutions, and Ecos Consulting.

The report can be downloaded free of charge at this link. Registration required.