Experts article
Innovation - data center measurement accuracy how to improve the energy efficiency
Keith Dunnavant, Munters & Anu Katka, Vaisala
Buildings and indoor air quality
Industrial measurement
In the next article, comes fromMunterstheKeith DunnavantAnd from visaAnu KatkaWill describe the future direction in the field of data center, and the effects of hvac measure for energy efficiency.With rising energy costs and governments are seeking to reduce greenhouse gas emissions, data center energy efficiency become the focus of international scrutiny.
The authors in data center energy management have long experience and abundant professional knowledge.MuntersIs for mission critical processes, including data center energy efficiency and sustainable climate control solutions for equipment manufacturers, and visa is the weather, environment and equipment manufacturers in the field of industrial measurement.
Energy consumption of data centres
"Power efficiency (PUE) as a" measure
The global demand for electricity is about 20000 terawatt hours;ICT (information and communication technology) industry to use 2000 TWH, data centers use about 200 TWH, accounting for 1% of the total.Therefore, the data center is an important part of energy consumption in most countries/regions.It is estimated that the global data center has more than 18 million servers.In addition to its own electricity demand, the IT equipment also need to form a complete set of infrastructure, such as cooling, power distribution, fire fighting, uninterruptible power supply, generator, etc.
To compare the energy efficiency of data center, the usual practice is to use "power efficiency (PUE) as a" measure.This concept is defined as a data center using the ratio of the total energy and the energy of IT use.Ideally the PUE is 1, which means that all the energy used to IT, and supporting infrastructure does not consume any energy.
To a greater extent, therefore, to reduce the PUE, the goal is to reduce the consumption of cooling and power distribution and other supporting infrastructure.Existing typical of traditional data center PUE is about 2, and large scale of data center can achieve 1.2 below.In 2020 the global average of about 1.67.This means that, on average, 40% of the total energy consumption is IT consumes.PUE is a ratio, however, so IT can't reflect the information on the total energy consumption, this means that if the energy level is higher than IT equipment cooling system, the PUE look very good.Therefore, to measure the total power consumption and efficiency and life cycle of IT equipment is also very important.In addition, from the point of view of environment, the consumption of power generation way, water should also be considered (including power generation and the cooling) and whether the use of waste heat.
PUE concept was first put forward by the Green Grid in 2006, and published in 2016 as the ISO standards.Green Grid is an open industry alliance, by data center operators, cloud providers, technology and equipment suppliers, the facility of the architect and the end user, to raise global data center ecosystem energy and resource efficiency, efforts to reduce carbon emissions.
PUE the commonly used method is still the computing data center energy efficiency.For example, in Munters, PUE is according to the peak value of each project and evaluation of annualized basis.When calculating the PUE index, IT is only considered in the PUE calculation load and cooling load.This is known as a part of the PUE (pPUE) or mechanical PUE (PUEM).Peak pPUE using electrical engineer to determine the maximum load and the size of the backup generator.Annual pPUE to evaluate the typical power consumption of a year, and comparing with other cooling scheme.While PUE may not be a perfect tool, but as the WUE (water using efficiency), CUE (carbon efficiency), and other measures and SPUE server (PUE) and TUE (PUE) can enhance the PUE correlation method is adopted, PUE have more and more support.
Data center trends
In the past decade, efficient vlsi data center in the data center relative share of the total energy consumption has increased, and many of the low efficiency of traditional data center has been closed.As a result, the total energy consumption has not increased significantly.The new large scale data centers that are designed for efficiency.We know, however, due to the artificial intelligence, machine learning, automated, driverless cars such as the emergence of many new trends, the demand for information services and computer intensive applications will continue to grow.Therefore, data center energy demand is expected to increase, and increase the level of is the theme of the debate.In the ideal case, compared with the current needs, in 2030 the global data center energy consumption will be increased by three times, but eight times more likely to increase.The energy consumption prediction includes IT and non-it infrastructure.Most of IT from the cooling energy consumption, or, more precisely, from the server cooling, cooling costs alone can easily about 25% of the total annual energy costs and above.Cooling is the necessary condition to maintain IT functions, of course, this can be achieved by building system to optimize design and effective operation.
An important trend in the near future is the server rack power density increased, some even as high as 30 to 40 kw and above.According to the data center professionals association AFCOM study, data center status report in 2020 showed that the average density of rack jumped to 8.2 kW per rack;Compared with 7.3 kW in 2019 and 2018 of 7.2 kW.About 68% of respondents said the rack density has increased over the past three years.
The shift towards cloud computing undoubtedly promoted the development of vlsi and hosting data center.Historically, a 1 mw data center is designed to meet the needs of Banks, airlines or university, but now many institutions and companies are turning to very large scale and hosting data center facilities within the cloud services.With the growing demand, is becoming more and more high to the requirement of data speed, of course all of these data centers are in service in the mission critical applications, so the reliability of the infrastructure is very important.
People also more and more attention to the edge data center in order to reduce delay and also liquid cooling is used to deal with the high performance chip and reduce energy use.
Temperature and humidity control
Air-cooled data center cooling energy efficiency is one of the main consideration of thermal channel/cold temperature and humidity control.Unfortunately, many existing data center still mismanagement of temperature and humidity control system, lead to low energy efficiency.New data center construction, on the other hand, tend to attach too much importance to temperature and humidity control, which has made great contributions to performance improvement.
In many cases, the ideal supply air temperature should be between 24 ° C and 25.5 ° C.However, thermal channel and cold temperature difference between (delta T) is very important.Usually, the delta -t about 10 to 12 ° C, but usually in the data center design will aim at 14 ° C.Delta - T ascension will lead to two aspects of income, is needed to reduce cooling system's fan motor, energy consumption and increase the potential of the strategy to achieve energy saving heat dissipation.
To promote energy conservation is the use of outdoor air data center of the cooling process.Energy conservation can occur directly, outdoor air directly into the cooling system and transmission to the server (after appropriate air filter);Can also indirectly, recycling of data center of air through the air - air heat recovery unit discharged into the environment.It can reduce cost and improve efficiency and sustainability.But in order to keep the efficiency, we should try to reduce air side pressure drop due to filtering.Therefore, if the air in recycling within the data center and not bring in outside air, it should be able to reduce or eliminate the need for filtering.
Cooling and ventilation need precise control, it is important to deploy efficient fan, keep the structure of the micro positive pressure and control indoor humidity.Should be spatial dew point control fresh air system, for example, in a low enough level, so that the cooling spiral pipe can be carried out only cold, and don't have to deal with latent heat load (reduce the humidity of the air).
Heat removal system of the overall goal is to keep IT equipment in the best state, at the same time, as far as possible to reduce energy consumption.Humidity is low, for example, may increase the risk of static electricity, and high humidity can cause condensation, and pose a threat to electrical and metal equipment, increase the risk of failure and shorten the working life.Has been proved that high humidity level and combination of various environmental pollutants will accelerate the corrosion of various components within the server.
Cooling must be carried out to eliminate the heat generated by the IT equipment, in order to avoid overheating and prevent failure.According to some research, compared with high temperature stability, rapid temperature fluctuations, in fact, the harm of IT equipment is bigger, so from this perspective, the control circuit is very important.
New IT devices are usually able to run at higher temperatures, this means that you can improve the inlet temperature, and are more likely to realize natural cooling and energy saving.Outdoor air can be directly or indirectly (as above) used to cool the indoor air, evaporation or adiabatic cooling can further improve the efficiency of energy saving.As needless to water use of dry heat removal strategy becomes more popular, these energy saving technology has been widely used.Since heat extraction medium (gas or liquid) temperature, the effective use of data center heat potential also will increase, such as waste heat can be used for district heating network.In Helsinki, for example, Microsoft and Fortum is to develop a collection of waste heat energy group projects.Microsoft will use 100% emission-free electricity data center, at the same time the multi-server Fortum cooling process of heat transfer to the clean and district heating system is connected to family, service and business place.The data center of waste heat recovery facilities may be the largest facility of its kind in the world.
The importance of accurate monitoring
In many modern facilities, is expected to be 99.999% uptime.This means that every year there are only a few minutes of down time.As a result of the IT infrastructure to deal with data and processes is very important and valuable, good performance is necessary.
A key feature of the design of data center is to provide the correct temperature for the server, and that only in the case of a control system with precise sensors to achieve.Room is bigger, may be more difficult to monitor, because their space temperature are more likely to change, so should be equipped with a sufficient number of temperature sensors to monitor to ensure that all servers, it is very important.
Keith Dunnavant Munters
Some servers may distance cooling device, some are far away;Some may be on the bottom rack, others in a higher position, so there was the possibility of a three dimensional change.In addition to the deployment of a sufficient number of sensors, therefore, to ensure that the appropriate place in the whole server room cooling device and the air flow is also very important.Through proper design and monitoring combination, can effectively control cooling and airflow, so as to meet the required specification.
In order to evaluate the effects of different variables on the annual energy consumption, Munters in three different location system model is established, in three different control application model operation;Each system has a 1 mw ITE load data center:
- The benchmark case scenario, the design and supply temperature is 24 ° C, return air temperature of 35 ° C (delta T = 11 ° C).
Case 2, the supply and return air temperature decrease 1 ° C (delta T remain unchanged).
In case 3, only will return air temperature decrease 1 ° C (delta T reduce).
The results show that the climate is mild, the energy consumption of the three scenarios are also low.Scenario 2, according to the supply and return air temperature by one degree, energy consumption increased by 1% to 2%.The increase of energy consumption in scenario 3 is significant;Only will return air temperature decrease 1 ° C (reducing the delta - T), the energy consumption of the three locations are increased by 8% to 9%.Temperature of a small deviation can produce very big effect, that highlights the delta - T and the importance of accuracy of the sensor.
Regardless of what kind of cooling method, the control of hvac process in a reliable manner and indoor conditions are crucial.To achieve this goal, data center management needs to have accurate continuous measurement for reference, because the good control circuit is closely related with the quality of measurement.Therefore, the high quality sensor is effective control of hvac process and stability of the indoor environment of the driver.When installation, however, the reliability of the sensor specification is not necessarily can as long-term performance indicators.Sensor value lies in its whole life cycle, because of frequent maintenance requirements may require high cost, and, as the Munters models have shown that the accuracy of even a small deviation can also lead to soaring energy costs.
In most cases, the value of the data in the IT infrastructure is very high, and usually is the key task, therefore, if can lead to high maintenance cost or high value data at risk, the deployment of low-cost sensors is meaningless.So users should find durable measuring equipment, can provide a reliable, stable readings for a long time, because it is important to the reliability of the device lifetime.
Factors influence the selection of the sensor
1. The reliability
Sensors on the accuracy of the installation points is obviously important, but more important is that sensor can long-term stability to provide accurate readings.Considering the high value data centers and are usually relatively remote location, the service life of the transmitter should be far higher than normal.Therefore, manufacturers should have experience in this field, and has the reputation of reliable measurement in critical environments.The traceability of the calibration certificate to ensure normal operation of sensor before they go out, and verified the reliability of the means that can keep the accuracy level for a long time.
Anu Katka, Product Manager, Vaisala
2. Maintenance
Avoid using high maintenance requirements of sensors, not only because of the costs involved, but also because of this kind of sensor can lead to a higher risk of failure.In addition, as mentioned above, the accuracy of sensor drift or decline will lead to huge energy costs.Data centers have higher levels of request for uptime, this means monitoring equipment maintenance operation should not interfere with the operation of a data center.Therefore, such as visa of instrument with a replaceable measuring probe or module can better meet the needs of customers;Especially this kind of instrument can remove sensor and offline calibration.If replace the measuring probe or module, you should also update the calibration certificate, it is also important.Ideally, the tool should be able to use the instrument suppliers in the field maintenance operations, and the work should be performed as part of the regular maintenance plan.
3. The sustainability
From the point of view of the sensor, the new technology allows users to upgrade sensor measuring components only, without having to replace or discard the transmitter, which helps to avoid unnecessary waste.Should be considered when deciding to buy, the supplier's certificate of environment and sustainability.This will help keep the sustainable development extending down the supply chain, and a driving force for the enterprise formation.Sustainable development is the core of Munters and visa to pull.Munters, for example, more than 1.5 gw have been installed in the global data center cooling equipment: save energy equivalent to 2% of the annual energy consumption of Sweden.Visa to pull in the financial times "2022 European climate leader enterprise" in the top five in the list.The list contains a during 2015 to 2020, greenhouse gas emissions most European companies.
conclusion
Because billions of dollars worth of critical data in the data center for processing and storage, thus consumes large server must be kept in the ideal temperature and humidity conditions, in order to prevent downtime.At the same time, along with the rising energy costs, we have an urgent need to reduce greenhouse gas emissions, improve energy efficiency, reduce energy costs and take better PUE measures.The driving factors of this "perfect storm" suggests that accurate control and optimization of hvac process is very important.
About the author
Keith Dunnavant is Munters sales vice President, responsible for the company's data center operations in the americas.Anu Katka is visa product manager, in charge of visa, global hvac product areas and data center.
You took also like...
The professional articles
Heat and cooling energy costs for the data center
Therefore, consume large amounts of energy industry is the strict censorship.The following article expounds how to implement high cooling efficiency in energy intensive data center.The article also introduced how to accurately measure to realize significant cost savings and improve environmental performance.
webinar
Accurate temperature and humidity measuring impact on energy consumption of cooling system
Cooling to keep temperature and humidity in the ideal level to achieve reliable server running is crucial.Even if only small changes in the cooling temperature, may have made a significant impact on overall running costs of data center, in the online seminars, we discussed the hvac precise measurement in data center energy efficiency play an important role.
