In I would advocate electrical efficiency modelling for data centres. New technologies in a hybrid model go hand in hand with servers and mine different applications in a data centre, deploy specific tools to address legacy applications; this would be possible if a proper bench marking is done to segregate mission critical applications.
More than the design, a simple, more accurate efficiency model is described that provides a rational basis to identify and quantify waste in power and cooling equipment, says Kevin Hughes, Business Development Director, Cooling Line of Business-Asia, Pacific and Japan, Schneider Electric. To this effect, estimates of electrical losses are typically made by summing up the inefficiencies of various electrical devices, such as power and cooling equipment.
According to him, it is critical to understand what data centre efficiency is and the power consumption pattern in a data centre efficiency model, if any hybrid approach is taken to drive efficiency. Data centre efficiency is defined as the fraction of input power delivered to the IT load. In an efficiency model, waste is everything other than what has been defined as the systems useful output. Clearly, the data centers DCPI (data center physical infrastructure) does other useful things besides provide power to the IT load these are called secondary support. It could be argued that the useful output of these DCPI subsystems (cooling or lighting, for example) should also be considered part of the data centres useful output.
Data centres are not built to produce cooling or fire protection or any of the other good things DCPI accomplishes. While these DCPI outputs are extremely useful to the internal workings of the data centre in helping it to produce and protect its useful output (computing), they are not themselves useful output of the data centre; nor is there any reason to believe that they must consume electricity. Non-power-path DCPI activities should be considered a necessary evil in supporting the data centres computing-- therefore, in the data centre efficiency model they are considered waste that should be minimised to the extent possible. All should be considered fair game for alternative designs and new technologies to reduce overall power consumption in the data centre.
An improved model for overall data centre efficiency also depends on how accurately individual components, such as UPS, are modelled. Characterising power and cooling components by using a single efficiency value is the common method, but it is inadequate in real data centre installations. The actual efficiency of a component such as a UPS is not constant, but rather a function of the IT loads.
To conclude, conventional models of data centre efficiency typically overstate efficiency because they do not properly comprehend the degree to which equipment is oversized, nor do they comprehend reduction of efficiency at the reduced loads where most data centres operate, says Hughes. An improved model provides more accurate numeric values for data centre efficiency, as well as insights into where the losses go and how they can be reduced.
Add new comment