What consumes the most energy in a data center?

Introduction 

 

Data centers have revolutionised the way we use digital technology. Allowing firms to outsource their data storage and processing, data centers play a major role in the global economy. 

To picture data centers, think of an enormous building full of computers storing information and processing data 24/7. The major problem with this technology is its high energy consumption

In this article, our experts explain how to calculate a data center’s energy consumption.

 

A few figures on data centers

 

Because of the digitisation of our economy, companies have become increasingly reliant on data centers to manage and store their digital data, even when IT does not have much to do with their core business. 

When it comes to the actual implementation, these companies have the choice between two main types of data centers

  • Some firms have their own data centers and store their data locally.
  • Other firms use cloud computing and outsource their data storage process to specialised firms (or for website hosting for instance). 

Data centers have been exponentially growing in size and quantity because of the increasing dematerialisation of data and the general increase in internet traffic. Indeed, they now account for 4 to 5% of global energy consumption and 1 to 2% of global CO2 emissions

About half of these emissions come from the manufacturing of servers (by extracting rare metals) and the other half from their energy consumption.

 

Are data centers energy-intensive?

 

Although this figure is considerable, please note that the use of data centers (via cloud computing) allows firms to save energy compared to individual on-site systems. 

Indeed, large data storage firms (such as Google, Amazon, and OVH) use the same servers for several users, thus optimising their storage capacity. As such, firms mutualise their energy consumption from data storage and platform hosting. 

Furthermore, as shown on this graph, the global energy consumption from data centers has only slightly increased since 2010, while the global volume of data processed by these same centers has increased almost eightfold in the same period. Indeed, a data center’s energy consumption is not proportional to its workload

 

Where does a data center’s energy consumption come from?

 

Data centers consume considerable amounts of energy. There are two main energy consumption sources for a data center: the operation of computer servers and the use of cooling systems.

 

The operation of data center servers

 

A data center is an immense warehouse filled with servers connected to the Internet via a broadband solution. These computer servers use electricity to store data,host websites, and run intensive computations, in the same way as your own personal computer. 

This is the main source of electricity consumption in data centers, although it is not the only one.

 

The cooling of data centers

 

The second major data center energy consumption source is the cooling of the machines. You may have noticed that when used for too long (or for intense tasks, such as video games), your computer tends to heat up. 

The same process appears in data centers because servers are constantly running and outsource vast amounts of energy, transformed into heat

To avoid frying the servers (thus making them unusable), data centers have cooling systems to keep the servers at a cool temperature. Scientists recommend machine rooms should have a temperature of 18°C to 27°C.

 

What are the different ways of cooling a data center?

 

There are different types of cooling systems.

Air cooling systems

If the data center’s geographical conditions are appropriate, outside air can be used to cool data center rooms. As such, filtered air is distributed to the computer rooms by means of machines and fans.

 

Adiabatic cooling systems

Adiabatic systems cool hot air by passing it through water (the latter then evaporates). This keeps the incoming air cool and naturally lowers the temperature in the building. 

 

Hybrid cooling systems 

So-called “hybrid” cooling techniques combine the two systems given above (adiabatic and air cooling).

 

Liquid cooling systems

Liquid cooling systems are the most advanced technology and considerably improve a data center’s energy efficiency. This method cools the processors by directly submerging the servers in non-conductive liquids. 

According to the method chosen, the cooling system is more or less efficient and therefore consumes more or less electricity.

 

How is a data center’s energy efficiency measured?

 

A data center’s energy efficiency is measured using an index: the Power Usage Efficiency (PUE). A center’s PUE is calculated by comparing the proportion of energy used to run the servers to the proportion of energy used for the entire operation of the data center (e.g. cooling, lighting, security, etc.).

For instance, a data center consuming 100,000 kWh for its servers and 70,000 kWh for the rest has a 1.7 PUE. The closer the PUE is to 1, the more energy efficient the data center is.

The share of cooling in a data center’s energy consumption depends on the type of cooling system used. 

In “hyperscale” data centers (i.e. data centers belonging to large data firms such as Google or Amazon), cooling is highly optimised. As such, their PUE is close to 1.15. This means that cooling represents 13% of the data center’s total electricity consumption.

Large firms have a considerable economic interest in improving their PUE: their size means that it is often profitable for them to make massive investments in their cooling infrastructure to reduce energy costs.

In traditional corporate data centers (such as in banks, for instance), the PUE is often near 1.7. Cooling is less optimised and therefore accounts for almost 40% of the data center’s total electricity consumption.

 

How can a firm measure its carbon emissions related to data center use?

 

Although a data center’s total kWhs consumed are crucial to understanding a firm’s energy bill, it is not a very meaningful indicator from a carbon perspective. Indeed, other factors are to be taken into account when measuring a data center’s CO2 emissions. 

 

The data center’s geographical location

 

Depending on its location, a data center will not be responsible for the same amount of greenhouse gas emissions. The carbon intensity of the energy mix in the data center’s area must be taken into account, as explained in this article

Data centers usually operate by connecting to the local electricity network. This power network’s emissions (from producing electricity) must be taken into account when calculating a data center’s real emissions.

 

How to overcome power cuts

Servers must run continuously. To this extent, a data center cannot suffer from power cuts from the electricity network

Therefore, firms using data centers have established systems to deal with local electricity network power cuts. One method is to use back-up generators, but this is a particularly CO2-intensive method. Alternative means of power supply must therefore be taken into account to effectively account for a data center’s total carbon emissions.

 

Conclusion

As discussed in this article, servers consume the most energy in a data center, followed closely by cooling systems. Therefore, data centers encompass many sources of carbon emissions

Although giant tech and data storage firms have made considerable commitments to improve and reduce this impact, a carbon audit must be considered on a firm-wide basis.

The new legislation for firms in regards to greenhouse gas reporting requires firms to include scope 3 emissions in their calculations

As such, firms using decentralised data centers must now include them in their calculations. To this extent, carbometrix offers solutions to easily calculate the impact of data centers in a firm’s carbon footprint.