With the cost of electricity and datacentre power consumption exponentially rising, it’s no surprise industry is turning towards green initiatives. Organisations that deploy environmentally sensitive technologies are doing their part to reduce the impact on the environment. But at what cost? According to a recent Gartner report, U.S. companies spend as much as 10 percent of their total IT budget on power and cooling. With such a substantial amount spent on just power and cooling, datacentre operators are under a tremendous amount of pressure to make a solid business case for going green. Will the installation of green products reduce energy costs and lower total cost of ownership?
And what makes a datacentre green? Is the datacentre green just because the organisation states it is? Of the 61 billion kilowatt hours consumed in 2006, 11 billion kilowatt hours is attributed to uninterruptible power supply (UPS) losses as a result of underutilised or simply inefficient UPS systems. The vast majority of these losses can be reduced by as much as 86 percent in best case scenarios. This sort of savings would perk the ears of any CIO and CFO, but to realise these savings operators need to focus on designs that center on system efficiency, interoperability, right-sizing, and the use of renewable energy sources. These four factors serve as guidelines to achieving a green datacentre.
System level efficiency is critical in achieving a high overall efficiency and a number of energy efficient UPS systems are available on the market today. The challenge is to see past the 100 percent load efficiency number listed on the specification sheet and focus in on the partial load efficiency. To be fair, most UPS systems in an Uptime Institute classified tier IV facility run at loads of 40 percent or less due to redundancies and failover strategies. Electronic Power Research Institute (EPRI) research suggests the average load factor of UPS systems in the field is 37.8 percent which results in efficiencies as low as 81 percent. In contrast, an integrated flywheel UPS system is 98 percent efficient at 100 percent load and 96 percent efficient at a partial 40 percent load.
Similar to the HP POD (performance optimised datacentre), a pre configured, fully contained datacentre, power and cooling technology has evolved in much the same manner. Containerised power and cooling systems that include UPS, switchgear, standby generator, and chiller systems are nothing new. What is new is sustainable and highly efficient containerised power and cooling systems. The inherent Achilles heel of a containerised power system has been the chemical batteries. By nature of the design, a containerised power system is more often than not deployed outside, exposed to the environment. The telecom industry is painfully aware of the challenges with chemical batteries, fluctuating temperatures, and frequent discharge cycles.
The integrated flywheel based UPS system, on the other hand, exhibits no level of degradation as a result of the environment due to its ability to operate in a wide ambient temperature range (0 to 40 degrees Celsius). Additionally, a mechanical and dynamic system is far more predictable given its vast network of more than 150 telemetry points, providing an accurate picture of system performance for every flywheel revolution. At 98 percent efficiency at full load and 96 percent at a partial 40 percent load, it is far superior to most of its legacy peers.
Interoperability between systems is often an ignored area from an efficiency standpoint, but is paramount when ensuring the reliability of the system. The bath tub reliability curve receives significant scrutiny and rightly should given the mission critical nature of the datacentre. The mistake is often made to couple two or more efficient products together to create a system. However, the end result from an efficiency standpoint is much worse than where one started. For example, two 92 percent efficient systems connected in series equals (0.92 x 0.92) 85 percent!
The deployment cycle of a 10, 50 or even 100 megawatt datacentre can be 18 months or longer. Economies of scale on up front costs are achieved by constructing larger power systems. However, this approach often results in installed capacity that is in orders of magnitude larger than the demand or IT load. Partial loads result in very low system efficiencies, although close to full capacity (40 percent) will be reached at some point in time in the datacentre lifecycle.
Interoperability between the integrated flywheel UPS, switchgear, standby generator, and chiller is ensured by a pre-fabricated containerised system where design, quality control, testing, construction, and verification is done at the factory rather than onsite. This saves valuable time during the implementation and deployment of the system on site.
The ability to manage and optimise IT loads in smaller and isolated blocks makes up for a significant savings. In fact, Active Power studies show an accumulated savings of more than 2 million pounds over 10 years on efficiency improvements alone is realised when right-sizing infrastructure using a containerised and modular design approach. Right-sizing the infrastructure to the load has the most impact on physical infrastructure electrical consumption. The greatest challenge in right-sizing is that conventional brick and mortar build outs take time, a lot of time. By the time permits are pulled, constructed, verified, tested, and commissioned for the next phase in a datacentre, power demand would be equivalent to the following, third, fourth or fifth phase.
Right-sizing using a conventional approach is far less agile than standardised, rapidly deployed containerised power and cooling systems. Individual sizes can range from 200 kilowatt up to three megawatts depending on the extent and pace of the deployment. Rapid deployment is partly achieved through relief from construction permitting and avoiding potential labor and union issues, but generally attributed to the lack of brick and mortar.
Renewable energy sources
61 billion kilowatt hours cost approximately 9 billion pounds or about 15 pence per kilowatt hour. Reversing the rampant energy consumption trend and ultimately realising a reduction in kilowatt hours consumed may be great for the environment, but may not do anything to the 9 billion pounds spent. With oil prices at approximately 63 pounds per barrel, a reduction in cost per kilowatt hour is unlikely in the near future.
Coal power production will reduce dependence on foreign energy sources, but will become burdened with the costs associated with carbon capture and storage to reduce climate impact. Nuclear power is likely to become a more significant contributor to energy production, but its ability to do so with predictable cost remains to be seen. Public sentiment on the development of new nuclear power plants has not been tested in nearly thirty years and heightened security and environmental concerns associated with radioactive waste disposal will continue to put pressure on this industry. One thing for certain is future electricity costs will continue to rise. Renewable energy sources such as solar, wind, and hydro power are often purchased at a premium. However, fossil or nuclear fuels will reach a level where renewable energy sources will become attractive as a 100 percent replacement or, probably more likely, as a substitute in what would be peak-shaving.
Renewable energy sources play an important role in containerised system design. Coupled with a containerised datacentre, organisations help replace the dependence on fossil or nuclear fuels or as a partial substitute and/or peak shaving to reduce grid power consumption and cost per kilowatt hour.
A containerised power and cooling system utilising an integrated flywheel UPS at 98 percent efficiency combined with a standby generator and chiller plant cater to all four factors in achieving greenness in the datacentre – systems efficiency, interoperability, right-sizing, and the use of renewable energy sources.
Tags: Design & Facilities Management, Power & Cooling