Energy: it’s becoming a broken record in the data center industry. It’s a major concern of data center managers and operators, with energy prices rising (whether due to inflation or increasing demand) and with the appetite of power-hungry facilities growing constantly. To make matters worse, government regulations on energy consumption, production or both always seem just one short step away. In this complex situation, data center managers are seeking ways to decrease their consumption in almost any way they can, including by raising the operating temperature of their facilities. But do the temperature limits of manufacturer warranties on servers create an unnecessary roadblock to conservation, or are they really necessary to keeping the equipment running longer?
Warranties Versus Efficiency
To draw customers, server vendors offer warranties on their products. Naturally, however, these warranties require that the customer operate the equipment within certain (reasonable, one would hope) limitations—if you use your server as a cup holder, for instance, and accidentally fry it by giving it a swim, the vendor is unlikely to offer you a new one. Most folks understand the logic behind this approach.
A conflict arises, however, with regard to server inlet temperature. Generally (and within reason, of course), electronic equipment operates better and more efficiently at lower temperature, so companies that once kept their data centers quite cool all year round were unlikely to run afoul of their server warranties. But as energy has consumed a larger portion of the average IT budget, companies began increasing (rather than decreasing) operating temperatures. Eventually, of course, a facility can reach a temperature where the air at the server inlet exceeds manufacturer specifications, thus voiding the warranty.
At this point, companies find themselves in a predicament: which is more valuable, warranties on servers or money that could be saved by increasing the data center temperature? It all depends on whether manufacturer specifications are on par with reality.
The Vendor’s Perspective
Vendors are (believe it or not) in the business of making money. When a vendor offers a warranty on a product, you can bet it’s done enough research to ensure it won’t lose money on the deal. A warranty is nothing more than a form of insurance—just think of how insurance companies operate to guarantee that they make money over time. Thus, vendors have more incentive to specify conservative operating temperature ranges (i.e., more-moderate temperatures) than the equipment might actually be able to handle.
Is this approach a malicious and greedy scheme on the part of vendors? Probably not. They are simply trying to ensure they don’t lose money on the deal. (Does the customer express any remorse about buying products at fire sale prices, even if doing so hurts the vendor?) But this does create a conflict when it comes to data center energy efficiency, as eWeek Europe (“Vendors Are Blocking Data Centre Energy Cuts”) notes: “Evidence suggests that facilities could operate at up to 113F (45C) without damaging equipment, but this would render equipment warranties void, as manufacturers specify a conservative operating temperature for the products.”
Kevin Timmons, CTO of CyrusOne, believes that servers “can be operated in much harsher environments than is recommended by the manufacturers. The example I always use is to think about the typical household cable or DVR box. It's essentially a Linux server, and look at the environment in which it lives. It’s typically stuffed into a cabinet or closet filled with dust with no ventilation whatsoever. It’s designed with a plastic case that is far from optimized for airflow. And it just sits there and runs for years.” Although the data center is slightly different than this example (mostly in scale rather than in kind), Timmons still suggests that “the notion that servers are fragile devices that demand constant inlet temperatures of 70 degrees Fahrenheit is antiquated at this point.”
The Customer’s Perspective
The customer has every right to operate its servers beyond the temperatures specified by the manufactures—at the expense of the warranty, of course. Thus, the customer must choose between the value lost owing to a potentially voided warranty (the warranty adds cost to the server—it’s not a free service) and the money lost by operating the data center at an unnecessarily low temperature.
But which option is better from a fiscal standpoint? That depends on a number of factors, including energy prices, how much energy is conserved at a particular temperature and how far that temperature falls outside the manufacturer’s specified range. Timmons notes, however, that “if you run data centers at any scale, you know that you can save significant operating expense by raising temperatures and widening humidity ranges. We think that we can save approximately 20% of our yearly operating costs by raising temperatures from 70 to 76 degrees Fahrenheit, not to mention the environmental impacts of lower energy usage.”
Earlier in the year (February), the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) announced its third edition of its “Thermal Guidelines for Data Processing Environments,” which “will be...groundbreaking in that it will enable compressorless cooling (all cooling through economizers) in many applications,” according to an ASHRAE press release (“ASHRAE Works to Expand Datacom Environmental Classes”). Although temperature needs for different types of equipment vary, this new release puts the burden on manufacturers to offer warranties with higher specified operating temperatures.
Since the new ASHRAE guidelines essentially side with the customers in the warranty versus temperature debate, opportunities thus become available for vendors to gain market share by increasing specified operating temperature. The value for the customer in this case is in operating expense, which has eclipsed the capital expense for a server over its lifetime. The vendor can thus provide the same product backing while enabling lower energy consumption by the data center in what amounts to an overall cost reduction. “Many manufacturers are now widening their recommended temperature and humidity ranges in keeping with the new ASHRAE standards, so this is a very positive step,” Timmons said.
A much heralded example is Dell, which in July announced that it was offering integrated IT equipment warranted for operation according to ASHRAE’s updated standards. According to the company’s July 28, 2011, press release (“Dell Leads Shift to 'Chiller-Less' Data Centers with Fresh Air Technology”), “Tested, validated and warranted to operate within the highest current temperature and humidity guidelines issued by ASHRAE, the servers, storage and networking equipment of the Dell Fresh Air cooling solution are capable of short-term, excursion-based operation in temperatures up to 113 F (45 C). This is the highest temperature warrantied in mainstream servers in the industry.” The Dell product seeks to enable the use of free cooling throughout the year in some cases, greatly reducing cooling costs for the data center.
The End of the Debate
Customers can still complain that vendors are setting their temperature specifications too low as long as traditional cooling is still required. Once free cooling (which may not be 100% free, but is close to it relative to its alternative) meets the needs of servers all year long within manufacturer specifications, the complaints will no longer have any basis. But even if PUE reaches a perfect 1.0, energy consumption will still be a problem, as the IT equipment is still a power hog even ignoring the supporting infrastructure. At that point, efforts at increasing the efficiency of the processors and other elements of the IT infrastructure will take center stage. But there will likely always be applications in which equipment operates at high temperatures, which will lead to customers questioning whether vendors are playing games by keeping their specified operating temperature ranges low.