 Hello, Paris, hello, everyone. I hope you are having a great time. Are you having a great time? That's good to know, because I am having a terrible time. I caught this flu this weekend. I thought, how I will extend to this stage? 15 minutes talking to 180 meters all. And tonight, I could not sleep with fever, and I had these kind of hallucination dreams and so on, you know? And I got an idea there. And I will need your help. So I ask you that you, if you have things on your lap and so on, put them on a safe place under your chest. And first, I want to talk why I need this. I grew up in a kind of ugly suburb of a very beautiful capital city at the Atlantic. And my first contact to nature and my interaction with nature was when going to the coast to surf. And you have these moments when you are surfing, and you can be completely wasted, either because you partied the last days or because you were surfing the whole day. But when there is this wave coming, you get this adrenaline and this energy, because you want to catch this wave, and you definitely don't want to be caught by the wave. So what I'm going to ask you is that we make a big wave and show the power of the Q-Con in Paris. And for this, I will give you a signal on three. And of course, the first slide is going, you know how a wave works, no? So let's go. One, two, three. This was not the wave I would like to catch. Come on, I want energy. I need your energy and your impulse. And I also want to hear the overflow rooms and the people at home. So everyone needs to make. So again, one, two, three, go. Up to the back. Yeah. Thank you, thank you very much. So this is the start of the story. The story I'm going to tell you is precisely about the urgency of climate change. Already in the 90s, when I was surfing, we had already all these environmental issues coming up. But the politics were still not so far with it. We had the first meetings in the Rio de Janeiro. In 92, we had the first Kyoto Protocol with the first commitments to climate targets. But they were all kind of insufficient. Although science already knew that it's an urgent matter and requires multilevel action and engaging every individual and organization to achieve the targets that we need. And it was not difficult when a friend of mine contacted me one day and said, hey, we have this team at the Deutsche Bahn and we do Kubernetes and so on. And yeah, I loved this. But I also said, yeah, this is actually a company with a purpose because I love trains. I love train travel. And I like to, I think, sustainable mobility with trains is very important to achieve the climate targets. And our company also believes this. So we are pushing forward to more train transport for freight and passengers. And for achieving this, we obviously cannot just build infinitely more infrastructure. This is quite heavy infrastructure. We also need to optimize the infrastructure, make it more efficient, more modern. And this requires a highly degree of digitalization. But digitalization comes with an impact. And the impact is, according to the International Energy Agency, quite remarkable. We see that in the next three years, according to the last report, we will at least require the amount of electricity consumption of Sweden and, in the worst case scenario, the electricity consumption of Germany. So we are reaching an exponential consumption phase. And of course, there are approaches to reduce it. But we also want to approach that the Deutsche Bahn started a green digitalization initiative in 2022 to address this question. And it came with the support of CEOs and CIOs. But it immediately raised also a very strong grassroots movement from developers, from employers of the company that started raising a lot of initiatives to small initiatives on their free times, making small contributions to measure, for example, the emissions of websites, or to make some papers on how to reduce the impact on your home office work, all kinds of initiatives raising there. And we also had a project where we wanted to measure and to bring developers to be able to measure the impact of our workloads. And we started because we have a cloud-based approach. We started to look at what the cloud providers give us in terms of tools. And unfortunately, we did not find very helpful information because the tools have limitations. They are not able to assign to an application. So if I have on my accounts multiple applications, there's no tax support, I cannot distinguish it. The granularity is very low. So I have 50% EC2 and 20% everything else. It's not telling me much. It's not very actual. Sometimes you get the information three months after the data has been created. And it does not give you any information on the energy used. Why is this important? Because emissions, they are calculated also with compensation. And this means that you are not able to establish a casual relationship between the actions that you do on your all day and the effect on emissions. And to understand also the problem of compensation, I would like to ask, who is your fan of Doctor Who series? Yeah, quite some. Yeah, great science fiction series. I watch it with my daughters every evening almost. And we hope still to finish it before they leave home. But yeah, there's this episode where they suddenly wake up in the middle of London, and London is full of a forest. So this is, for me, a good metaphor of compensation. Of course, compensating carbon is an important measure. But we cannot think that with an exponential growth scenario of exponential growth of energy demands, that we will be able to compensate all this energy. Because there are effectively limits to growth, physical limits. And so we need to address the different strategies for sufficiency, what I really need, and why for efficiency. How do I minimize the resource use and production? And consistency, which would be the case, for example, of compensation. So focusing, first of all, on the efficient part and on the role of developers. Back to the agents on this talk. And the topic developer empowerment. So developers are effectively the everyday decision makers in what comes to software. We don't need management papers to say we are doing this and achieving this target. If we don't engage developers and give them the tools in the hand, we will not be able to make a change in the way we develop code, in the way we manage infrastructure. So we pose the question, what tools and approaches are able to empower these developers? Let's start with platforms. We have a platform strategy at the Deutsche Bahn. And in the last years, we converge the whole subsidiaries to enforce standardization in this level. And the importance of platforms on this aspect is that we can leverage effects and provide a high level of standardization and open to contributions that improve it continuously and provide the same design aspects that everyone can use without even thinking, am I doing it green or not? It's just helping you to be green per default. I myself, I'm a bit biased. I love Kubernetes. I was product owner of two teams in this field and have a personal affinity to this. But I believe Kubernetes is a tool for efficient cloud operations and for carbon footprint reduction excellence. It's not only a container orchestration tool. It's much more powerful than that. It's a platform building tool, and it's a green IT tool. We have, with our clusters, achieved one of the interesting very high density of containers. We have some shared clusters. And we had then put, of course, the node autoscaling. And we managed to achieve around 70% of CPU utilization on these shared clusters. It's quite a high utilization. So you would say, well, we are effective. We don't need to know anything else. The fact is, however, only 15% of this utilization was effectively used by the applications. We could see this, for example, on Grafana dashboards, where there are a lot of reserved CPU utilization, but only a very small part of it is being used. Why does this happen? Well, once it runs, no one cares. We have a lot of things to do in our business all day. So I define requested limits. Traditionally, the API requests limits 512 megabytes. It is running. I make it conservative, because I think maybe there will be a lot of people coming, and then I cannot burst. So rather have it conservative. And it stays there. And it's taking CPU. It's occupying the node and wasting resources. Fortunately, Kubernetes offers us vertical pod autoscaler. There are also, of course, horizontal pod autoscaler and so on. But I find this one particularly interesting, because it is taking out of the end of the developer the need to think what would be the appropriate resources and rather making recommendations or even automatically adjusting the container workloads to be optimized to their needs. So let's go out with expensive get work and make sure to use vertical pod autoscaler. The second aspect related to the sufficiency is scheduling. And Kubernetes is also good on this. You know, we have all these worker rights movements that manage that we are able to work eight hours a day and have eight hours to do whatever we want and eight hours to sleep. I don't think most of you do this anyway. But let's assume everyone works from nine to five, like would be the traditional approach. This means two thirds of the day, there's no one at the office. And your workloads are running there. Development, environments, test environments, we're testing it, we're testing stuff at midnight. Just turn it off, use cube downscaler, annotate your deployments, so simple like that. It will turn off when you leave the office, it will turn on again when you come back and it's working. You don't even need to care for it, it does it every day, it saves a lot of energy and it costs nothing. Last aspect, we need information, visibility. It's like if I am traveling by train and my train is late or it fails, what do I do if I did not have good data informing me, providing me how will be my alternatives? So this is where we came with Kepler. We wanted to grab the energy metrics. And I will not go into detail in Kepler here because there are at least 50 people in this room for sure that knowing detail how Kepler evaluates energy statistics and transforms this into watts. But the interesting there is that you can actually get information from high level up to the container level. So really have an information on near real time on how much energy your single components are using. And you know in Grafana, it's also possible, you can make for example an annotation from your pipeline and associate a certain change in your code with an impact on the consumption. We tried to roll it out, we thought it's actually yeah, great open source, it's available, let's just roll it out. It was not so easy. The first one was the enterprise readiness in terms of security. There were some unneeded dependencies there that we needed to remove. We were able to give a contribution which was great. And the second one was, well, there were a lot of metrics. So there was someone being called in the night on the on-call duty and you know how people hate this. No one likes to be called in the night for because the monitoring system is overloaded. So we had to reduce the scraping interval which was by default on three seconds to much less and actually we don't want to keep also so much data. So it's enough if we have 10, 20 seconds scraping to give this information about what is the effect of the different commits. So data is the source of everything and we have seen on the left side where you see the green is the Kubernetes pushing back to Grafana and it allowed us to make these beautiful dashboards. Dashboards are always something beautiful. I find them at least and I think many people love dashboards and with this we had the first tool that really brings this in the hand of developers to make a change. And then we of course started getting more and more people interested on this and to gather the whole cloud carbon footprint data and we required a middle layer with our data lake with a data set with data governance that is in the middle and allows for other systems to couple and also create information for other potential agents. Like for example, architecture management. An enterprise architecture management, we see here an example of the dashboards that we are building with our enterprise architecture management system. And so it's about really bringing the tools also and the information to the tools that are already being used by these people. So as a final remarks, I would like to say the first important thing is start. Start now, it is urgent. The issue is urgent and we are all in an early stage of knowing how to actually measure energy. We have lack of information. We don't know if this data is good but it's better as no data. Some information is better as no information. So just start, you will see how many people because really developers have an intrinsic motivation not really to save costs but everyone has an intrinsic motivation to care for our planet, I believe it. So just start small and beautiful and you will see how your community and your system around the platforms and around green IT will grow. And focus on empowerment, focus on the agents, on the people that are on the all day taking decisions with their actions. Encourage developer initiative and their intrinsic motivation. So thank you very much. Let's stick together if you want, contact me or any of my colleagues that are also on this conference and it was a pleasure to be here and I would like to ask the moderator to come here and because I want you also to experience this wave. So let's do it one more time if you don't mind. So one, two, three. Thank you very much. Have a nice day. Thank you so much, Walter.