Data Centers are Going Green

Going “green” by reducing electricity consumption in data centers certainly has the attention of manufacturers of information technology equipment (ITE) and of data center operators. Cost and availability of electricity are just two of the factors driving interest. A globally increased emphasis on reducing greenhouse gases is another key factor. Energy efficiency initiatives will continue to be implemented, and few would argue against their benefits. Could some of these initiatives contribute to a possible increase in the number of electrostatic discharge (ESD) events or in their severity?

Going Green in Data Centers

The past few years has seen an increased interest in improving the efficiency of how computer data centers operate. The basic economics of reducing operating costs is one clear factor driving this interest. Another compelling reason is a desire to operate data centers in a more environmentally sustainable manner. It has been estimated that improving energy efficiency on all data centers worldwide by 20% could be equivalent to eliminating the need for some 150 250-MW coal-fired electric power plants. [1]

To be sure, IT equipment manufacturers and data center operators are taking notice and implementing innovations to reduce electricity consumption and related greenhouse gas emissions.

While no one would deny the importance of this activity, some beneficial actions could have unintended consequences for electromagnetic compatibility (EMC) practitioners, including the potential to increase the number of equipment malfunctions¹ caused by ESD.

What possible connection exists between improving energy efficiency and ESD? Let’s take a look at how data centers are kept cool and some of the ways we can make that process more efficient.

Data Center Cooling Techniques

In a traditional data center, the heating, ventilating and air-conditioning (HVAC) infrastructure used to keep the room cool uses nearly as much electricity as the computing equipment itself. Various analyses indicate approximately 45% of electricity used is to power the ITE and 55% for HVAC and related infrastructure. Whatever the specific numbers, the message is clear: the cost to keep the IT equipment in its operating temperature and humidity range is very similar to the cost of operating that same IT equipment. Implementing techniques that reduce the cost of cooling these facilities can have a very large impact on overall cost of operation and the electricity used.

ITE manufacturers generally specify the temperature and humidity ranges in which their equipment is to be operated. More often than not, these environmental conditions are established to provide a reasonable assurance of quality, reliability and longevity of the equipment. If the operating temperature is too high, the electronics under the covers could overheat, causing a variety of undesired conditions such as operational errors, component damage and, in extreme situations, smoke or fire. If the humidity level becomes too high, several undesirable conditions can happen, including condensation which can create electrical shorts or arcing, and growth of mold or mildew.

A limit on minimum humidity is often established to reduce the occurrence and severity of ESD events to the ITE in the data center. Typical lower limits for relative humidity (RH) have traditionally been in the neighborhood of 30%. For example, the 2004 guidelines from the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) [2] include an allowable operating class with RH maintained no lower than 20% and a recommended operating class with RH maintained no lower than 40%. The lower limits on RH for controlling ESD are often based on experience and may be on the conservative side, particularly for 24/7 operations with business-critical applications.

The design characteristics of electronic equipment need to be aligned with the conditions of the environment in which it will be installed and operated. Focusing solely on ESD, if the equipment is designed to meet the ESD test limits of the international standard CISPR 24 [3], then the operating environmental conditions should be within a range that can be expected to prevent discharges that are more intense than those represented by the 4 kV contact discharge and 8 kV air discharge limits of that test standard. Conditions likely to create discharges more severe than these test limits should only be allowed for equipment that is designed to withstand more severe discharges.

Traditional approaches to control temperature and humidity use computer room air conditioning units (CRACs) distributed around the perimeter of the room. CRACS are quite good at lowering the temperature and humidity of a room, although locating them near the perimeter can create temperature differences throughout the space. Maintaining humidity above some lower limit, however, generally requires some type of humidification technique used in conjunction with the dehumidification provided by the cooling coils of the CRAC. The complex system of dehumidification for cooling and humidification for meeting a lower limit on humidity impacts the infrastructure and operating costs of the overall HVAC system for a data center. Given the realities of electricity consumption, alternate strategies to reduce demand and cost are being considered.

Alternative cooling strategies include:

• liquid cooling of equipment
• alternate air flow paths through equipment
• localized cooling of equipment in the data center
• wider operating ranges for temperature and humidity
• air-side economizers
• water-side economizers.

All five of these strategies provide real benefits in terms of reducing how much electricity is needed to operate the environmental infrastructure of a data center. No single one of these strategies is ideally suited to all situations, and each comes with its own particular benefits and challenges.

Let’s focus on the last three.

The first of these three, wider operating ranges for temperature and humidity, is a rather straight forward technique. The more tightly controlled the environmental conditions are in any facility, including data centers, the more complex the infrastructure is. If I want the temperature in my house to stay below 25 C in the summer, I simply set the thermostat for my air conditioner to that temperature. If the temperature drops down to 22 C, my operating condition is still met. No further control or action is needed. However, if I want to keep the temperature at 25 ±2 C, then my home infrastructure will need to operate the air conditioning when the indoor temperature is above 27C and the heat source if the temperature drops below 23 C, as may happen at night. This latter arrangement, with its smaller operating range, would require a more complex control system and cost more to operate. The same concept applies to data centers but on a large scale where, in addition to temperature, humidity levels are also controlled. The wider the operating ranges, the less complex the HVAC system and the less costly it is to operate (in theory at least).

The last two strategies, air- and water-side economizers, are sometimes called free cooling. These techniques take advantage of the ambient conditions outside the data center to supplement or replace traditional techniques for temperature and humidity control.

A very simplified description of an air-side economizer is a system that directly utilizes outside air to provide cooling instead of traditional air conditioning. When ambient conditions are favorable, cool outside air is brought into the data center and warmer air from the data center is exhausted to the outside. A similarly simplified description of a water-side economizer is a system that uses a body of water to provide pre-cooling of the chilled water used to lower the temperature inside the data center. More complete descriptions of these forms of free cooling can be found in multiple sources, including [4] and [5].

To maximize the effectiveness of these cooling methods, wide enough ranges for operating temperature and humidity need to be allowed. Recent activity with ASHRAE has aimed to do just that: allow and encourage operators of data centers to increase their operating ranges for temperature and humidity. One key result of this work is a revision to ASHRAE’s guidelines for temperature and humidity in data centers, [6] which have been incorporated into the 2011 edition of the European Union’s code of conduct for data centers. [7] The revised guidelines from ASHRAE includes an environmental class that allows moisture levels down to 0.5ºC dew point without special precautions, and relative humidity as low as 8% if appropriate static control measures are implemented. These changes represent a significant reduction in the humidity level in data centers. The additional static control measures are intended to alleviate concerns over potential increase in ESD events as a result of reduced humidity.

Relationship Between Humidity and ESD

Conventional wisdom and years of experience indicate that allowing the humidity level to become too low is likely to increase the number and severity of machine malfunctions from ESD. Manufacturers of ITE and operators of data centers are aware of numerous situations where equipment malfunctions caused by ESD happened at an increased rate during times when humidity levels were low. For decades, ITE manufacturers have been resolving problems caused by ESD to their products by raising the humidity in the operating environment. Those of us who live in cold-weather
climates know all too well that ESD events above the threshold of human sensation tend to happen more often when humidity is low.

The observation that noticeable ESD events happen more often when humidity is low does not tell us what specific mechanism is impacted by humidity. What is quite apparent is that if humidity is too low, equipment malfunctions can happen. Keeping in mind the need for compatible equipment design characteristics and environmental conditions, quantifying how low humidity can become before ESD-induced malfunctions increase in frequency would be beneficial. Now, let’s examine the relationships between humidity and an ESD event, including the mechanism by which ESD causes electronic equipment to malfunction.

A static electricity discharge is a complex event. The critical elements of the event include:

• charge of one object or surface with respect to another
• pre-discharge voltage differential
• arc through which transient discharge current is delivered
• discharge current, including its rise time, amplitude and derivative with respect to time.

Does variation in humidity affect any of these elements? If it does, are they elements that will make discharges occur more often or be more severe when they do happen? What is it that makes a discharge more severe: a higher pre-discharge voltage, the amplitude of the discharge current, the rise time of the discharge current, or perhaps a secondary discharge?

The answers to these questions are not simple. We know that all of the indicated elements of a discharge have the ability to make a discharge more severe. Which ones are important is determined by the characteristics of the electronics being impacted. Some types of circuits are sensitive to high-frequency disturbances and will be more affected by ESD currents with faster rise times. Other circuits are not so sensitive to high-frequency energy but are affected by the amplitude of the disturbing current. For these circuits, the peak amplitude of the current is the more important element of a discharge. Other interactions and sensitivities also exist for different circuit types.

Research by Pommerenke [8] has demonstrated a clear correlation between the arc length of the discharge current and both the current’s rise time and peak amplitude. An arc is a non-linear circuit element that is affected by humidity. When arc length increases, the rise time of the discharge current increases and the peak current amplitude decreases. Over the range of 700 to 1200 µm for arc length, rise time varies from approximately 0.5 to 2 ns. Over the same range of arc lengths, peak current for a 5 kV discharge varies from 25 A down to 10 A. The demonstrated variations in these two critical elements of an ESD event with change in arc length are a good indication that the impact of humidity on arc length cannot be ignored.

A recent Green Grid paper by Swenson and Kinnear [9] provides an interesting, indepth discussion about charge generation and discharges. It includes a description of some research into the relationships between humidity and charge generation. One conclusion of that work is that humidity can indeed affect the static voltage that is generated on a surface. This work also indicates the possible affect of humidity on ESD should be considered when altering specifications on humidity in data centers.

Summary

Humidity does directly or indirectly affect all of the critical elements of an ESD event listed above. It has a direct impact on the static charging of surfaces and on the length of the arc through which the discharge current is delivered. Static charging of surfaces is a critical factor that determines pre-discharge voltage differential at the point of the ESD event. Arc length affects the rise time and peak amplitude of the discharge current. Caution should be exercised when allowing reduced humidity in data centers based on these findings, since they are preliminary and reliable quantitative data are not yet available.

This finding does not mean that going “green” in data centers will necessarily create additional equipment malfunctions because of ESD. It definitely does not imply that the industry should stop pursuing relaxed temperature and humidity operating ranges as part of plans to reduce energy consumption and cost in data centers. It does tell us that EMC professionals can expect additional challenges as data centers reduce energy use. It also points to the importance of close interaction between EMC and HVAC professionals, as well as the many other engineering disciplines involved in designing and operating ITE and data centers.

Additional work is needed to quantify the challenges and risks associated with allowing lower humidity levels in data centers. The industry will be well served by knowing how low is too low. Until that level of detail is discovered, the approach included in ASHRAE’s revised guidelines [6] is probably a prudent one: accept a wider range of humidity levels to allow the introduction of effective energy efficiency initiatives and require additional static control measures if humidity levels are allowed to drop below a point that is considered “too low” to coexist with the design point of mainstream ITE.

ASHRAE is planning a research project to help quantify the relationship between humidity and the severity of the reactions of electronic equipment to ESD events. Members of the EMC community are participating in this research effort. We can expect to hear more about this activity in the future.

References

1. www.usctcgateway.net.
2. Thermal Guidelines for Data Processing Environments, Atlanta, GA, ASHRAE, 2004.
3. CISPR 24: 2010 Information technology equipment – Immunity characteristics – Limits and methods of measurement, CISPR, 2010.
4. Air-side Economizers – 42U, http://www.42u.com/cooling/air-side-economizers.htm.
5. Water-side Economizers – 42U, http://www.42u.com/cooling/water-side-economizers.htm.
6. 2008 ASHRAE Environmental Guidelines for Datacom Equipment, ASHRAE, 2008.
7. 2011 Best Practices for the EU Code of Conduct on Data Centres, http://re.jrc.ec.europa.eu/energyefficiency/pdf/CoC/Best%20Practices%20v3.0.1.pdf.
8. Chundru, R., Pommerenke, D., Wang, K., Van Doern, T., Centola, F.P., Huang, J.S. “Characterization of Human Metal ESD Reference Discharge Event and Correlation of Generator Parameters to Failure Levels – Part 1: Reference Event,” IEEE Transactions. Electromagnetic Compatibility vol. 46: 498 – 504. November 2004.
9. Swenson, D. and Kinnear, J.T., The Role of Relative Humidityand Dew Point on Electrostatic Charge Generation and Discharge (ESD), The Green Grid, 2009, http://www.thegreengrid.org.

Note

1. The phrase “equipment malfunction” is used herein to describe unexpected, and unwanted, departure from normal performance of electrical and electronic equipment when it is in operation. This includes temporary degradation of performance or loss of function that is automatically corrected by the equipment or that requires action by the operator to correct, and physical damage to the equipment or its circuitry.

John Maas is Corporate Program Manager for EMC for IBM Corporation and has responsibility for IBM’s worldwide EMC regulatory compliance program. He has over 25 Years of EMC experience including hardware design and test. He has been involved in international standardization for much of his career and currently is active in IEC SC77B/WG10 and the US advisory groups for IEC TC77, SC77A and SC77B and CISPR/I. Mr. Maas can be reached at johnmaas@us.ibm.com.