The evolution of four grades to PUE 2 is fully improved PUE

The measurement of power usage efficiency (PUE) provides a common standard for data center contrast and improvement of power usage. But this standard provides a lot of interpretation space, many organizations will use their own way to explain, in this article, we will analyze the potential defects of PUE, and the PUE 2 How to make up for these flaws to do some explaining.

Pue defects

Pue is a means of assessing and tracking power utilization in the datacenter, which helps us understand the effects of the facility's improvements, but we quickly get bogged down in puehttp://www.aliyun.com/zixun/aggregation/31080.html " > Extreme Sports Competition, it can be said that many manufacturers claim that the PUE value is not credible, because no one vendor can provide details on how to obtain these values or how to measure, so Pue has become a manipulated number.

October 29, 2009, Green Grid released the Pue white Paper version 2.1, which is now often cited Pue v2, which provides a more accurate reporting of pue, but still confusion, May 17, 2011, Data Center Efficiency Working Group Center efficiency Task FORCE,DCETF published a paper that describes the recommended methods for measuring and reporting pue.

Almost everyone knows that the definition of Pue is very simple, and it equals the power delivered to the datacenter divided by the actual power that it equipment uses.

Pue= total power to data center/IT load power

Pue if equal to 1.0 is the best state, this means that all input power is used by IT equipment, there is no refrigeration, no lighting, no wiring, because they will always consume a part of the power, technically it is impossible, most traditional data center Pue value between 2.5 to 3.5, but the newly built data center and container structure data center Pue usually claims to be less than 1.1, and if so, it's perfect.

So what makes the value of Pue so low? Simply, reducing the numerator or increasing the denominator can make pue smaller, the smaller the pue the better, there have been signs of human manipulation, such as the choice of the best time to measure, the cold outdoors, the lighting system all closed, the user measuring almost no online, or even shutting down redundant refrigeration systems to measure, The Pue value of this time is certainly very low, but it is indeed far from the truth.

From the equation given above, we know that data center input total power is smaller, pue may be better, do not know how you measure it load power, if it is read UPS output power, then you have to add PDU (distribution unit) transmission loss, wiring loss, cabinet fan electricity, etc. So that the denominator in the equation gets bigger, make pue smaller, if your data center's power supply is shared by other functional facilities in the building, such as offices, meeting rooms, restaurants and lobbies, or shared air-conditioning system refrigerators or cooling towers, if you cannot measure independently, you need to accurately estimate the data center input power, This is critical to computing pue.

PUE 2 Improvement

In simple terms, Pue was not originally intended to compare the data center differences, it is only a trace of the independent data center cycle time of the energy situation, manipulating this number for the data center itself is not a benefit, and get accurate pue value is only for internal improvement, Green Grid has consistently insisted that pue should not be used as the index of comparisons between data centers, but how to prevent people from comparing pue curiosity, so that caused the false pue value of the rampage, in any case, even if you want to do comparisons, there should be a unified basis, at least the measurement standards, measurement methods should be consistent

PUE v2 has made 4 major changes to the PUE index.

PUE v2 established different types of PUE measurement methods, Green grid divided it into 1, 2, 33 levels, respectively, corresponding to the basic level, intermediate and advanced, DCETF added a more basic level 4th, PUE0 to PUE3 to identify, the higher the level of the more stringent, Next I will focus on the DCETF definition of pue0,pue1,pue2 and PUE3.

Except for Pue 0, all measurements must be converted into energy, means to introduce a time component, such as a kilowatt hour to replace KW, which must include all input energy, such as the heat generated by gas combustion must be converted to kilowatt hour, in addition, all measurements must be measured at a specified point in time, at least for a continuous measurement of one year.

PUE v2 now stipulates that the energy consumed by the office, NOC, etc. must also be included in the PUE calculation formula.

PUE v2 also clarified that 17815.html "> renewable energy, such as wind and solar energy, has no impact on PUE, and PUE is completely unrelated to where energy comes from and what energy it is."

Here's a summary of the four different points of Pue V2, understanding the minimum requirements for each level, and why you shouldn't compare them with Pue.

PUE0: The equivalent of the original PUE, it is still measured at discrete points in the data center of the total input power and it load power, the main improvement is now it is required to use IT equipment, the value of the peak period of utilization, PUE0 can not show the impact of dynamic load, so the timing is different, The final result is a big deviation. In addition, it load power using UPS output will increase non-it overhead, such as PDU loss, cabinet fan loss, etc. PUE0 is still based on power rather than energy, so it can only be used in 100% power-driven data centers and there is no way to calibrate different fuels. In short, it is useful to track the effects of various changes in independent data centers, but PUE0 should not be used for comparisons between different data centers, even if they are designed to be the same.

PUE1: This is the first level of energy or power consumption based on the new Pue, which requires the total kilowatt hour in the last 12 months, all fuel types generated energy will be converted to a common value, that is, kw per hour, compared to PUE0, this is the most important improvement, But it still follows the UPS output as it load energy, so it is possible to commit PUE0 computing it load the same error, according to the actual situation, these errors will affect the value of pue to varying degrees. If there is no PDU, no cabinet fan, and everything else is measured in accordance with the PUE1 method, the final calculated Pue value should be very close to the real value.

The only difference between pue2:pue1 and PUE1 is the measurement of it load energy, PUE2 is the output of the aggregated PDU, as mentioned earlier, if the difference between pdu,pue1 and PUE2 is small.

PUE3: This is the most accurate pue measurement method, and eventually all data centers are eager to adopt this measure, it needs to accurately measure the input power of each IT equipment within 12 months, currently only a few data centers have the ability to do this step, but this is undoubtedly the most accurate method. In my opinion, any pue that claims to be below 1.3 is nonsense unless all facilities are built in accordance with the PUE3 level.

Now we should know exactly how to measure the value of Pue, when comparing the values of PUE1, PUE2 or PUE3, we should keep the measurement points consistent, so that they are comparable, in fact they are also the relationship between apples and oranges.

PUE v2 eliminates ambiguity, it provides several different methods of measurement, making it possible for most people to compute their own PUE values.

PUE V2 also established 4 levels for the PUE value, and a full range of measurements should be included in the design for new or reconstructed data centers, so that the data center can truly become an energy-efficient facility.