Do-IT-Yourself: the Power of Data Center Self-Assessments

Do-IT-Yourself: the Power of Data Center Self-Assessments

It's a fact: The power and cooling costs of data centers are rapidly rising. In 2006, the total amount of power used by data centers represented approximately 1.5 percent of total U.S. electricity consumption, according to the U.S. EPA's 2006 data center energy efficiency report [PDF]. The cost totaled $4.5 billion, about as much as was spent by 5.8 million average households for the year.

The growth in computing demand, and technology's ability to accommodate that growth, is really what's driving the power and cooling issue. Some companies now spend more on data center utility bills than they spend to acquire brand new systems. And at most companies, data center energy consumption is taking an increasingly larger bite out of annual facilities budgets, making this a high-priority issue for the CIO.

If you are hoping to optimize the efficiency of your data center and cut power and cooling costs, don't remodel your data center or build a whole new one just yet. This step-by-step guide to assessing and analyzing data center costs can help you, the IT manager, strategize to maximize efficiency in the data center you already have.

Getting Started

Step One: Tie off with the facility manager and business strategists. Historically, the IT manager has addressed the data center power and cooling issue from a technology and power capacity perspective, while the facility manager has focused on managing the costs of the physical data center, including its cooling capabilities.

For the organization to understand the bigger issue surrounding power and cooling, getting on the same page as the facility manager is the first step. Together, you can bring data center cost and efficiency concerns to an array of top executives, presenting a strong business case for optimizing the data center.

Furthermore, understanding the needs of the business by meeting with the business executives will allow you to understand future requirements of the data center more clearly. Business strategists can discuss with you what the company's strategies are to gain market share and be competitive within the industry, and therefore predict computing demand.

Step Two: Assess your resources. Once you and the facility manager are on the same page and you understand the needs of the business, it's important to decide whether you have the resources to do your own data center assessment or whether a professional assessment is the right choice for you.

In general, you will need the staff or specialists to assist in evaluating three main areas that factor into a thermal assessment: Hardware, Racking Practice, and Air Flow Management.

Further, consider the cost benefit of doing the assessment yourself vs. hiring an expert. Professionals are trained and experienced in doing quick data center assessments, or more in-depth assessments, in short time periods, while it may take a company's less experienced IT staff much longer to do a thorough assessment.

Data center assessments can take advantage of sophisticated modeling technologies to help companies achieve greater energy efficiency and cost savings in their data centers, as well as anticipating how data center power and cooling needs will continue to grow and change. One solution HP offers is a 3D Thermal Zone Mapping, which provides a multi-colored graphical thermal map of the entire data center. Using data from the model generated from its proprietary software, HP provides customers with a comprehensive report they can use to structure the data center for optimum efficiency and savings.

Armed with this data, the CIO can use the information to:
  • Understand the thermal characteristics of a data center.
  • Identify overlapping regions of cooling capacity. This is useful in showing locations in the data center where high-density and/or mission-critical equipment could be placed for optimal cooling.
  • Analyze "what-if" scenario changes in the data center, such as room layout or infrastructure changes, cooling capacity changes or failures.
  • Ensure that a proper cooling solution has been achieved by addressing areas of over-provisioning and smoothing out isolated hot-spots.
A sample image of the type of map data center managers can develop through basic thermal assessment

Performing the Assessments

Once you have determined that you have the necessary staff expertise to proceed, you can begin to assess any of the three main areas, Hardware, Racking Practice, or Air Flow Management, depending on your current strategies and environment. Here we'll address hardware first, followed by racking strategy and air flow management.

Hardware. For this part of the assessment you should examine how your hardware is being utilized. Ask the following questions:
  • Is hardware at maximum productivity and being fully utilized?
  • What is the performance of servers per watt of power?
  • Are cooling capabilities coming into the data center fully optimized?
In order to determine this information, you can use power management software that measures the amount of power used vs. the actual demand. One additional piece organizations should keep in mind is how much power they are allotted from the local utility company.

In addition, data center managers should begin or continue projects for server and storage consolidation. For example, take a look at migrating legacy applications to eliminate older servers that can be replaced with new, more energy-efficient servers. This can be a continuation of an existing blade strategy, or the impetus for putting one in place.

Upgrading hardware can result in up to 70 percent in infrastructure energy efficiency. In addition to servers and storage, data center managers should consider improving transformers and uninterruptible power supplies, efficiency chillers, fans and pumps.

Racking Practice

How your server racks are aligned and where and how the racks distribute hot air is pertinent to the efficiency of your data center cooling. Analyze the layout of your racks in conjunction with your airflow measurements. It may be helpful to chart or draw out a model of your data center complete with temperature measurements in each area. This may help determine why hotter air is being pushed to some areas and why certain racks are being cooled more efficiently.

While assessing your racking practice, ask yourself the following questions:
  • How is airflow being distributed to the racks and is the layout of the racks affecting airflow?
  • Are blanking panels in use to prevent the mixing of hot and cold air in the racks?
A white paper that discusses racking practices and other considerations, "Cost Model for Planning, Development and Operation of a Data Center," is available from HP.

Airflow Management

Analyzing the airflow within the data centers will determine if air is being circulated optimally to maximize the use of cooling technology.

With a do-it-yourself assessment, you can take basic inflow and outflow temperature readings and conduct a self-generated design analysis to reflect best practice. There are some market-available applications allowing you to interpret your findings. With the results of a do-it-yourself assessment, you will be able to understand the basic cooling patterns in your data center and strategize for a more efficient setup, as long as you feel confident in your measurements and how you've approached the issue. You need to consider this and the time it would take in the context of a professional analysis.

If you are looking for a complete analysis of your data center's cooling patterns, you may want to consider a professional assessment. Effectively measuring airflow in the data center can be an extremely complex and time-consuming process. Experienced professionals have the education and resources to perform in-depth testing, including thermal dynamic measurements, and can also provide a trustworthy, comprehensive report on data center airflow, mixing patterns and recommend potential Computer Room Air Conditioner (CRAC) adjustments.

However, if you choose to do your own airflow analysis, Lawrence Berkeley National Laboratory hosts a Data Center Energy Management website that addresses the topic. A comprehensive self-benchmarking guide is also available for download from the site.

A common practice for determining airflow patterns is to place heat sensors throughout the data center, including at tile level, under the floor, at mid-section of the servers, at the top of server racks and above racks, to take temperature measurements. Once you have these measurements, ask yourself the following questions:
  • Is airflow circulating throughout the data center, including underneath the floors and above the ceilings in the most efficient manner?
  • What obstructions to airflow exist (piping, cabling, air conditioning ducts, construction columns, etc.) that may be creating non-uniform airflow?
  • How is hot air re-circulated throughout the data center?
Best Practices for Maintaining an Optimized Data Center

Once you've completed your assessments and made adjustments in Hardware, Racking Practice and/or Airflow Management, you've made great strides in transforming your data center. Consider the following best practices as you continue to evolve the efficient operation.

Move from Room Cooling to Row or Rack Cooling: Anyone who has ever stepped foot into a data center can describe the experience as chilling Standard practices entail keeping data centers within a temperature range of 68 to 75 degrees Fahrenheit (20 to 24 degrees Celsius) for system reliability.

However, many organizations don't assess airflow patterns of the data center and end up over-cooling the entire room, which is not an efficient use of cooling power. Through the airflow assessment, you should determine that the supply temperature can be up to ten degrees higher than originally expected, since redirected air flows will use the cool air more economically. If air conditioners are specifically integrated with rows of racks or racks themselves, cooling will be more efficient. Background and references on this topic are available in an HP technology brief, "Data Center Cooling Strategies." [PDF]

Hot Aisles/Cold Aisles: Data centers should be configured in such a way that hot air travels through one aisle of racks as cold air flows through another. Servers should be situated back-to-back so hot air being pushed out is not blown into another server in close proximity. Air conditioning units should be oriented perpendicular to hot aisles. The use of blanking panels can help to prevent hot air from re-circulating to the front of the racks. Furthermore, using longer rows can minimize the penetration of hot aisle air in cold aisles.

Eliminate Gaps and Cable Cut-Outs: Gaps in rows of servers and between floor panels can cause hot air leakage into cold aisles. Furthermore, cut-outs in the floor for cables can leak cold air to hot aisles. Eliminating gaps and sealing cable cut-outs will help to prevent inefficient cooling.

Consider Economizers: Air-side economizers include a sensor that allows outside air to enter the data center when temperature and humidity conditions are appropriate. While there is some concern that outside contaminants will cause electronics to wear or shut down, the data center's filtration system will typically prevent contamination.

Raised Floors: Floors that are .8m to 1.0m high in the data center will improve airflow distribution under the floors.

Efficient Hardware: Getting rid of unused hardware and replacing old servers with energy efficient servers can boost efficiency. Also, enabling power management on all of the active servers will improve efficiency.

Joe Bottazzi is the vice president and general manager for HP Services' Americas Technology Services. For more information on data center energy efficiency solutions, visit HP's Data Center Services site.