The Shape Of Things To Come For The Data Center Industry, 2nd Edition
A look back and forward at the data center industry.
Syska presented the first edition of this subject at an Uptime Institute meeting in New York City in early 2009, and made some fearless predictions for data centers in the next five to 10 years. Since nearly six years have passed, we think it’s high time to review whether our 2009 predictions were accurate and to venture some new predictions.
2009 Prediction 1
• Prediction: Liquid cooling of self-contained cabinets and computer equipment will not become common.
• Reasons: We felt that self-contained cabinets were too expensive and offered no operating economies or increased reliability over air cooling. The market didn’t seem ready for liquid cooling of computer equipment. Purchasers still remembered the mantra that water above the data hall raised floor was to be too risky.
• Accuracy: Correct
2009 Prediction 2
• Prediction: Air cooling of cabinets and computer equipment will remain the norm.
• Reasons: We felt that equipment load densities were still low enough to make legacy air cooling the method of choice going forward. Plus, the major equipment suppliers had a vested interest in retaining the legacy technology and the marketing clout to back it up.
• Accuracy: Correct
2009 Prediction 3
• Prediction: ASHRAE TC9.9 acceptable entering air thermal envelope will get much larger.
• Reasons: To be transparent, we had two staff members attending the ASHRAE TC9.9 meetings at the time and we knew the manufacturers were working on expanding the allowable service conditions for their equipment to respond to energy efficiency and sustainability concerns. As a matter of fact, while this article goes to press, another round of modifications are being made to the thermal envelope. The “Recommended Range” will likely expand even more, and the lower end of the minimum dew point will be relaxed (meaning humidification may not be required at all, unless a nominal amount may be required for human comfort).
• Accuracy: Correct
2009 Prediction 4
• Prediction: Airside economizers and Kyoto Cooling will become more common.
• Reasons: We felt that energy efficiency and sustainability concerns would drive the market toward more efficient air cooling solutions.
• Accuracy: A straight statement of accuracy is complex on this issue. The acceptance of airside economization has increased over the years, though lately there has been some concern over the need to use sophisticated humidity control (whether it is to humidify or dehumidify). Even more importantly, the users of outside air economizers have expressed concern over the issue of indoor air quality as they affect the IT equipment. This has been especially true in industrialized areas and areas with marine or desert conditions. The use of Kyoto Cooling may see some growth for those customers for whom the outside air economizer is not a viable option. With a few minor modifications, a Kyoto Cooling system can be made to perform almost, if not as well as, an indirect evaporative cooler. The jury is still deliberating on this concept.
2009 Prediction 5
• Prediction: Once 95°F entering air temperature is acceptable for computer equipment, there will be no need for refrigeration equipment.
• Accuracy: Correct, but adoption is slowed by the cost and availability of suitable computer equipment. Some manufacturers provide products for higher entering air temperature, but the products have a premium cost and market adoption has been slow. Also, ASHRAE TC9.9-2011 added expanded allowable temperature ranges and a method to estimate the effect of increased entering air temperatures on product reliability. We have seen some enterprise users increase temperatures into the allowable range but, in general, users other than enterprise have been reluctant to take the risk.
2009 Prediction 6
• Prediction: There will be more emphasis on UPS efficiency and acceptable technology will increase it to 98%.
• Accuracy: Correct, 98% efficiency is attainable today with Active Power and S&C Purewave product in normal operation and with other manufacturer’s products in offline mode (often termed eco-mode or econo-mode). The Active Power product has a premium cost to other technologies and many potential users have been reluctant to pay the premium unless high cost of purchased electricity will produce an acceptable rate of return with the Active Power product. The S&C Purewave product is a medium voltage product and has some unique features that have not been readily accepted by many users. The offline mode feature is often purchased but less often enabled, since many users are reluctant to accept that it would always perform as advertised.
2009 Prediction 7
• Prediction: 400Y230 volt distribution to computer equipment will become more common until 277 volt power supplies become available. Then 480Y277 volt distribution will become common.
• Accuracy: Correct on 400Y230 volt distribution becoming more common; not correct on 480Y277 volt distribution. 400Y230V distribution has become more common but it has not superseded legacy 208Y120 volt distribution. High available fault current levels on 400Y230 volt systems have proved troublesome, particularly downstream of large capacity UPS systems. Another recognized problem is the need to employ 4 pole circuit breakers (at added cost) for 400Y230 volt systems where 3 pole circuit breakers had sufficed for legacy systems. Computer equipment power supply manufacturers produce 277 volt power supplies but these products bear a premium cost and are not available in significant quantities. The market seems to have stayed with the 100 to 240 volt power supply that can be used for either 400Y230 volts or 208Y120 volt systems, and 277 volt power supplies seem a sideline like 48 volt DC power supplies.
2009 Prediction 8
• Prediction: Virtualization will become more common and flexible.
• Accuracy: Correct. No need for any lengthy discussion here. Many users have eliminated thousands or even tens of thousands of servers. One major financial institution has a lower critical load now than six years ago.
Overall, we are pretty much on the mark six years after we made the 2009 predictions and are confident our accuracy will hold up during the next four years.
Making predictions is risky for the predictor and often provides great laughter for the reader. Ask Bill Gates who famously stated, “640K ought to be enough for anybody.” Nevertheless, we now offer some new fearless predictions for the next five to 10 years:
• 2015 Prediction 1: IT and network improvements will be able to cover over the inevitable reliability cracks in the electrical and mechanical infrastructure.
• 2015 Prediction 2: As a result of 2015 Prediction 1, fewer data centers with high cost Tiers III/3 and IV/4 electrical and mechanical infrastructures will be built.
• 2015 Prediction 3: Large scale cloud computing providers will continue to build massive data center capacity while reducing their user fees. This will make life difficult for retail colocation providers and challenging for wholesale colocation providers. The situation will change after these large scale providers obtain a large market share and then begin to raise prices, resulting in government anti-trust intervention.
• 2015 Prediction 4: UPS operating efficiency will not increase beyond 98%. Improvement beyond 98% will come at a high cost and will be employed only where the cost of purchased electricity is high.
• 2015 Prediction 5: The percentage of DRUPS installations in the US will increase as owners strive to shrink the cost of electrical infrastructure and as more DRUPS providers enter the U.S. market.
• 2015 Prediction 6: Direct liquid-cooled computer hardware will become common for high-density computing. Densities of 200kW/cabinet will not be unusual.
• 2015 Prediction 7: Use of traditional CRAC/CRAH technology will dwindle as more data centers employ direct/indirect airside economizer and adiabatic cooling to reduce Total Cost of Ownership.
• 2015 Prediction 8: Water cooled centrifugal chillers will no longer be used in new data centers.
In summary, we reviewed our predictions of six years ago and found they were pretty accurate. Then we made some new predictions for the next five to 10 years. Shall we schedule an appointment to review them in 2020 and apply our 20-20 hindsight?