Good software is green

Greenhouse gases contribute to a continuous increase in the earth’s temperature. The reduction of greenhouse gases is therefore necessary, even if they are caused by the development and operation of software. Software in itself only indirectly causes the release of greenhouse gases. However, it can also be used to minimise the production of greenhouse gases in other areas, for example by using software to monitor and improve other processes.

There are already promising mathematical models that computationally approximate the energy efficiency of various functions in an application and already deliver quite good key figures. The use of AI-based systems has also provided good results for evaluating the energy efficiency of software – although it should be noted that neural networks themselves usually cause higher power consumption than classic calculation models. There are already various tools for developers to measure the energy efficiency of their code. However, the results of many methods vary greatly based on how, in which time periods and in which context the energy consumption is considered. This makes it difficult to compare digital products with each other, but they can still provide a good indication of where an application has potential for improvement.

Electricity in a local grid is made up of various sources. Large parts are still generated from fossil fuels. The share that is generated by renewable energy sources depends strongly on the geographical location and the prevailing energy policy attitude there. Rarely do institutions have much influence on how the electricity they consume is generated. Most of the time it is not even clearly traceable. However, this problem can already be counteracted by avoiding server locations when operating an application where it is already known that large parts of the local network are fed by non-environmentally friendly sources.

A constant conflict of software development is the urge to stay at the cutting edge of technology. However, the environmental impact of a physical server is not only limited to the power consumption during its lifetime, but already comes into play during the production of a computer, as well as later during its disposal. The environmental impact of its manufacture and disposal can even be higher than that of the energy required during operation. It would therefore make sense to keep servers in operation as long as possible. In principle, computers do not wear out because they have no moving parts. Nevertheless, there is an expiry date: for example, when they can no longer cope with modern workloads, they need to be replaced. Software that is compatible with older equipment helps to extend the life of hardware and thereby reduce its environmental impact.

At the same time, the relationship between utilisation and energy consumption is not proportional, also because a running computer always has a certain basic consumption. The better a computer is utilised, the more energy-efficiently it functions. Servers can only reduce their energy consumption to a certain extent when the load is low, so they can respond immediately to requests when they are made. Ongoing server scaling can save energy during periods of low load by running software on fewer physical machines. Serverless” architecture also helps keep power consumption to a minimum during idle periods.

A more invisible environmental impact is the internet itself. The internet consists of a worldwide network of devices that consume additional energy. Every request over the internet causes energy consumption. How large this is depends mainly on the distance between client and server, how many network devices have to forward the request, how they are operated, etc. – but also on the protocol used.

Resource-efficient software should therefore avoid unnecessary data transfer and choose the appropriate protocols to transfer data as efficiently as possible.

Demand shaping deals with the practice of adjusting the service, or the quality of the service, inversely to the demand. Video conferencing software, for example, reduces video quality and prioritises audio transmission during busy periods. This methodology can be used to make applications more environmentally friendly.

For example, the user experience can be adjusted when the cost of operation becomes too high. Since there is sometimes a cost in usability here, an “eco-mode” can also be offered as a conscious decision during use.

It is not necessary to weigh up all eventualities from the outset. Individual conscious decisions can already make a big difference. Above all, it is important to observe developments and continuously introduce improvements step by step. This approach is not only in the spirit of sustainability, but also ensures adaptable software with proactive maintenance at the same time.