Sustainability in Software Engineering, Part 3: Targeting Data Transfer

Transporting data over the network consumes energy and thus causes CO2Emissions. If you want to contain this by reducing energy consumption, you must not lose sight of the data traffic over the network.

Their tasks give software developers unimagined opportunities to shape the future. Rather, it is their responsibility to ensure a better tomorrow. A new column on heise Developer by Martin Lippert.

If you consider the energy consumption of data traffic, two variables are decisive: the amount of data transferred and the distance that the data has to cover. Two principles can be derived from this:

  1. The less data the software has to transport over the network, the less CO2Emissions arise.
  2. The shorter the transmission path, the lower the CO emissions2.

Since the exact energy consumption of data transmission in the network and in the Internet depends on numerous parameters, it varies greatly. In addition to the parameters of quantity and distance, the actual energy consumption also depends in detail on the specific network nodes involved and their energy efficiency. The lines between the nodes and their exact technology also play a role.

To determine the effect of the amount of data, I use the following basis: 1 Mbyte of data transmitted on the Internet requires approximately 0.0023 kWh of electricity. The value comes from the Microsoft online course on the subject of “Sustainable Software Engineering”, which in turn refers to a study by “The Shift Project”. It is a rough average. Therefore, from my point of view, it is not suitable for calculating or deriving an exact value for a specific case study, but it gives a rough impression of the magnitude.

Based on this, 1 GByte of transmitted data requires approximately 2.3552 kWh. Converted to CO2Emissions results in around 1.22 kg of CO2.

I took a look at my house connection: A look into the wireless router revealed almost 850 GB of data to be transferred in a month. That translates to around 1034 kg of CO2 – in a month. As a reminder: the long-haul flight, which was used as a comparison in the second part of the series, causes around 1883 kg of CO2 per flight in economy class for one passenger.

A reader pointed out to me in the previous part of the series of articles that I can only do the long-haul flight with 640 kg of CO2-Emissions instead of using the total emissions value. This criticism is justified. Emission values ​​are often used for comparison in “CO2-equivalent “amounts converted. Therefore it makes little sense to use pure CO2Emissions on the one hand with CO2– Compare equivalent amounts on the other hand. The corrected value for long-haul flights is 1883 kg of CO2-equivalent emissions instead of 640 kg of pure CO2Emissions. Thanks for the feedback!

It is therefore worthwhile to look at and analyze the network traffic of your own software in the data center. Considerable amounts are accumulated every month – both in the internal network and in external traffic.

For web applications, it is worth taking a look at the book by Tom Greenwood “Sustainable Web Design” [1]which lists numerous ways to reduce the amount of transferred data for web applications, in some cases drastically. A positive side effect, by the way, is that the browser usually loads web applications optimized in this way much faster.

What probably doesn’t play a big role for in-house software can make a big difference for software that runs in the public cloud. The approaches to keep the data transmission route short are not all new. Companies should

  • carefully select the region in the public cloud and optimize the proximity to the users,
  • rely on content delivery networks to keep frequently required data close to the users instead of transporting them over long distances on the Internet and
  • Use browser cache settings to reload data and content as rarely as possible.

Here again applies that many optimizations not only the CO2-Reduce emissions, but at the same time accelerate the software by loading less and faster data.

The Green-Cloud-Computing-Studie on the subject of video streaming gives a good impression of how different technologies perform in terms of CO2 emissions:

  • Glass fiber (2g CO2)
  • VDSL (4g CO2)
  • 5G cellular network (5g CO2)
  • 4G cellular network (13g CO2)
  • 3G cellular network (90g CO2)

The concrete figures, which in this case relate to an hour of video streaming in HD, are not very meaningful, but the relationship to each other is interesting. If you have the option of transmitting data via a VDSL connection and WLAN, you should use it instead of using the cellular network. 5G is the preferred choice for cellular networks, as it comes close to the energy consumption of a VDSL connection.

To home page