Energy consumption of SVT Play
When you watch a video on SVT Play some amount of electricity will be consumed by your device, the network and SVTs servers. This energy consumption will in turn lead to some carbon dioxide emissions. The carbon dioxide emissions for watching one video is not so much if you compare it to for example eating a hamburger meal or driving a car. However with 1.3 million active users on SVT Play every day and more than 4.2 million video streams started every day it adds up to a considerable amount.
When the user starts a video a lot of data is sent from SVT servers to the users TV, computer or cell phone. This data will pass through network equipment, radio base stations, routers in your home etc. Throughout this entire process there is equipment that consumes electricity. The amount of energy needed to play a program depends on several different things. A TV for example consumes more energy than a cell phone, mobile networks consume more than optical fiber networks etc. A rule of thumb is however that the more data that is being sent, the more energy will be consumed. Sending a video stream requires a lot of data compared to for example an image or a web page. On average 1,5 petabytes of data is transferred from our servers every day. This is quite a lot and it requires a lot of energy.
So what can we do to minimize the energy consumption for SVT Play?
The thing we can do that will have the most impact on energy consumption is to minimize the amount of data that needs to be sent over the network. To do that we put a lot of effort into investigating the latest and most efficient technologies.
Most of the data that is sent over the network from our servers consists of video streams, so adjustments to those will have the largest impact.
An example of a rather new technology that we are implementing now is the video encoding format HEVC (High Efficiency Video Coding). This technology can in theory enable up to 50% size reduction compared to the previous format, especially for higher resolutions. With HEVC we can increase the quality of the video stream with the same amount of bandwidth. Or we can deliver a video stream of the same quality using less bandwidth.
Another optimization that we do is to encode different types of videos with different settings. For example animated content can be more compressed than other types of videos.
While the absolute majority of data consists of video streams, we also do optimizations to other kinds of data. For example we optimize the images that are used to describe our content by investigating and using the most efficient image formats. Here we can achieve a 30–50% size reduction of the images.
Often these optimizations will also improve the user experience. For example the videos will start quicker and have less interruptions. Also many users have a limited data plan on their mobile subscription, so they will be happier as well.
At the same time people are buying new TVs with bigger screens and higher resolutions. This creates a demand for higher quality video streams which in turn requires more data. Streaming of 4K resolution demands much more data than traditional HD resolution.
One idea that we have been thinking about is to show the user how much carbon dioxide emission the different choices of streaming qualities produce. Just like some restaurants show you the carbon footprint of the different items on the menu, we could provide similar hints to the user and let them choose depending on their current needs.
Finally we can conclude with the amount of users that we have, even small savings can have a big impact. We try to optimize on many different levels while balancing the user needs. Often the optimizations we do to improve the user experience, also will minimize the energy consumption required.
Here you can listen to our talk at the IVA-conference Science & Society Forum: Växande IKT-sektor och fler datacenter — hur påverkas elförsörjningen? 5th of November 2020: https://www.iva.se/event/science_and_society_forum_2020/
Hilda Stenberg and Andreas Bjärlestam
