What if we unveiled the often overlooked environmental impact of data centers together ?

Travel with Gaël Duez to Sweden to meet Life Cycle Management expert Stanislava and to France to meet Benoît, co-founder of Hubblo and NGO Boavizta who works on impact evaluation and energy/material efficiency for businesses.

Join us for an eye-opening episode on how data centers can play a crucial role in building a sustainable digital future.

Together, we explore:

✅ The definition of data centers and their various types, including hyperscalers, enterprise data centers, and colocation data centers.

✅ Top tips and insights on how to make your data center operations more eco-friendly and reduce energy consumption.

✅ The automation and democratization of impact evaluation.

✅ The controversial topic of... cloud!

❤️ Subscribe, follow, like, or stay connected in your own way so you never miss an episode!

Learn more about our guests and connect

Stanislava Borisová has a strong background in Life Cycle Assessment, with a Master of Science in Industrial Ecology and extensive experience in assessing the environmental impact of various sectors.

After joining the Research Institutes of Sweden, she focused her attention on studying the environmental impact of data centers. She now works as an Expert in Life Cycle at IVL and stays involved in the data center industry through her participation in the Sustainable Digital Infrastructure Alliance (SDIA), among other initiatives.

Benoît Petit, co-founder of both Hubblo and the NGO Boavizta, has nearly a decade of experience as a Cloud and SysAdmin engineer. His work is all open-data and open-source, reflecting his values. Benoît's background is primarily in IT infrastructure and cloud, and he founded Hubblo two years ago. The company helps businesses and communities reduce the environmental impact of ICT through impact evaluation and energy and material efficiency.

Stanislava’s LinkedIn
Benoît's LinkedIn

📧 You can also send us an email at greenio@duez.com to share your feedback and suggest future guests or topics.

Stanislava and Benoît’s sources and other references mentioned in this episode

Hubblo
Boavizta’ studies (in FR but easily translatable):

And its API & tools: https://boavizta.org/tools

RISE Research Institutes of Sweden

The article that Gael mentioned that is not an article but a podcast :

Environment Variables episode discussing the report assessing 7.2 million data centers world-wide

Öko-Institut
Fraunhofer-Gesellschaft

CEDaCI project
"PCRs under development" - Electronic devices, components and services
GitHub

Blauer Engel

Cloud scanner (project in Boavizta)

GitLab CI : evaluation of impact

CICD tool (related to the SDIA)

FinOps : public cloud management discipline
SDIA and its softawere
David Mytton’s blog

Gauthier Roussilhe, researcher in France
Françoise Berthoud and her article Le numérique, espoir pour la transition écologique ? 🇫🇷
Jean Baptiste Fressoz’s video on "transitions énergétiques"

Books :

The Best of Times, the Worst of Times Futures from Frontiers of Climate Science by Paul Behrens

Towards the Energy of the Future, written by KTH, the University in Stockholm

GeSI
GSMA

Transcript

Gael: You're listening to Green IO , the podcast for responsible technologists, making our digital world greener, one byte at a time. I'm your host Gael Duez, and I invite you to meet a wide range of guests working in the tech industry to help you better understand and make sense of its sustainability issues, and find inspiration to positively impact our digital world.

If you like the podcast, please give it five stars and a nice review on Apple, Spotify, or your favorite platform to spread the word to the millions of people working in the digital sector and now enjoy the show.

Hello everyone. In this episode, we go to both France to meet Benoît and Sweden from where we welcome Stanislava to talk about sustainability in the data centers industry. When I first met Benoît in Paris at a cafe having a nice lunch under a cool summer day. I know it's cliche, but it's true. He told me a sentence, which has stuck in my mind since now on. Does anyone actually read the figures in these reports?

And he did so because A) I was actually guilty at that time of overconsuming reports on Green IT, field, which I had just discovered, and not paying attention enough to the data and the methodology underlying them. And B), I told myself, whoa, this guy knows what he talks about. And indeed, Benoît having cofounded, both Hubblo and the NGO , after almost a decade working as a Cloud and SysAdmin engineer, knows what he talks about.

It's actually pretty easy to double check because everything he produces is open data and open source, two values he cherishes. But we'll come back to this point later on. Stan was introduced to me thanks to Chris Adams, the director of the Green Web Foundation, when we were discussing life cycle assessment. His word, "she is one of the most knowledgeable on this topic".

Which makes total sense knowing that after her Master of Science in Industrial Ecology, she joined the Research Institute of Sweden to study, well, LCA applied to data centers. She has now moved toward an expert lifecycle position at IVL, but stays close to the data center field via her involvement in the SDIA, the Sustainable Digital Infrastructure Alliance, among other things.

Welcome, Stani and Benoît. Thanks a lot for joining Green IO today.

Stanislava: Hi Gael. Thanks for having us.

Benoit: Hi.

Gael: So before jumping into the nitty gritty of data centers environmental impact, my first question is always about your personal journey in the sustainability area. So Stani, how did you become interested in sustainability and in the sustainability of our digital sector in the first place?

Stanislava: So yeah, thank you for asking. I think one of the most powerful experiences that actually led me to join sustainability field was an exposition that I went to with my high school. And I remember hearing about what issues we're facing and how possibly people in 50 years or so would have to ration water because there would not be enough clean water anymore.

And I remember that being very scary and very strange to think of, and potentially that was too powerful to do something about at that time. So I sort of let it be and I started studying international business instead. But as I was going through all the courses on accounting and finance and marketing, I thought there was something missing to that and I thought there had to be something more.

So that's why I actually selected to follow an elective on sustainable business. And at some point I started to think to myself, yes, this is very interesting. This is how things should go. But at the same time, how do I know that what the companies are saying is actually true? How can I verify that they actually are sustainable and it's not just some claims?

So that then led me to the masters that you've already mentioned in the beginning, industrial ecology, where I learned how to quantify this sort of business behavior and impact. And after a few years of working as a consultant in IE in Industrial Ecology, I joined RISE Research Institutes of Sweden, and that's how I got introduced to the data center industry.

It was maybe a bit of a coincidence and a happy accident because I was not very involved in that before, but I had worked in technology so there was an interest, and after that I've realized that there is so much, it's such a huge heterogeneous industry, and so much can be done regarding sustainability on so many different levels, and I'm happy to be still involved.

Gael: So that's funny because it all started with water, and actually I believe water is a topic we will discuss regarding sustainability in the data center industry. And what about you Benoît, how did you come involved in sustainability and all this work you've done in green IT ?

Benoit: I can't remember about a very specific event that drove me there. I think I've worked in IT like many, many other people, not really realizing what was behind in terms of impacts, because I didn't realize at first that our modern society's had such an impact on the environment and that it was really a problem for the future.

So I realized that piece by piece very progressively and I changed many things in my personal life so I could feel aligned with what we do for the future. And so I started working on my own thing, actually, I was working on cloud infrastructures at that time, and I felt, okay, if I have to start somewhere, I should start on my own problems, my own impact.

So first, how is it big? How could I evaluate this impact? And my first answer was like, "that seemed like super complicated". So I found work from the area, which is a research lab in France. I found some projects about it, but nothing that I could use directly in my own context. So I don't really know why, but rather than staying in the company, I pretty quickly thought that I had to leave and work on this topic as much as I could. And so I started developing a software for measuring the power consumption of servers, of ID servers. Then I happened to be discussing with people who were building an NGO from scratch.

So I started discussing with them, and then later I realized that there was some demand for the work I've done on on the software. And then I said to myself, "okay, maybe there could be a business here, so I could work on this topic 100% of my time". So I don't have to find a new job in like one year.

And so you have Hubblo, and that's how I came to the topic.

Gael: So synchronizing a bit more what you've already started to do in your personal life with what you can do and the expertise you already had in your professional life.

Benoit: That was the idea.

Gael: Okay. And you know, I spotted very recently a discussion in the climate action tech community about an article that I must admit, I forgot the title, who had quite a lot of debate about the number of data centers worldwide. It was stated that we have already 7.2 millions data centers worldwide and three millions in the US. Some people finding it quite consistent, some people finding it completely crazy. This number. And then it connected me with one of the first topic we discussed with Stani a few months ago actually, which was her ontology.

Let's define properly, what is data center, what is a hyperscaler, what is an enterprise data center, colocation data center, et cetera. And I would love Stani, if you could help us set the stage a bit regarding the data center landscape. What are we talking about? What are the main numbers and how would you actually explain this controversy?

Regarding the number of data centers, and I will not enter into the other controversy about the number of servers because I think it will take half a day to talk about this one.

Stanislava: Well, I'll try to do my best. Indeed. It's very hard to measure and to count how many data centers there are because you can have a data center at your own company, which is a very small closet with a few servers. And technically that counts as a data center. But to sort of distinguish between the main groups of data centers, and of course you can complement what I say.

I would say that you have enterprise data centers, which are the data centers owned by the companies themselves to have their own data.

Then, there are colocation data centers, which are such data centers that external companies own. And then if you have the need to store your data somewhere, but you don't want to do it in-house, you rent some space and then you just populated with your servers and then you store your data there.

And finally you have the so-called hyperscalers, which are the large companies such as Google, Amazon, Facebook, and Microsoft that own their own data centers, even develop their own servers and technology, rent out parts and are just very big, which is why they're called hyperscalers.

Gael: And so depending the definition, the 7.2 millions could make sense or not. I mean, if we pitch ourselves that a data center being, there's massive facilities that you can see from the sky which belongs to Google or Amazon. Obviously there is not 7.2 millions worldwide, but if you include what you've labeled enterprise data centers and, and colocation data centers, do you believe such a number could be correct?

Stanislava: Yeah, I think it's possible, especially if we think of all the small data centers, one rack or two racks, I can imagine it could be that much.

Gael: Okay, so let's now enter the main topic of our episode today, which is sustainability in data centers. For the sake of clarity, let's put aside cloud for the moment. I mean, obviously everything that we're going to start discussing regarding enterprise data centers or colocation data centers will at some point apply to your cloud service providers.

However, you don't interact directly with metrics or the electricity consumption, et cetera.

But truth is there are still millions of companies interacting one way or the other with data centers and not having migrated to the public cloud or private cloud, whatever. And my question would be, "what can you concretely do when you are in charge of a data center?"

You are head of infrastructure, you're obviously a CTO, CIO. How can you run a greener data center and maybe, I think Stani, once again, you told me once about PUE fatigue, that you were a bit fed up with everything focusing only on PUE. So could you elaborate a bit on what would be your main advices, your main insights on how to green your data center operations?

Stanislava: Sure. So maybe just to mention about this. Tiredness of PUE. I think my main reason for not liking it so much, it's because it's become such a marketing tool and it's very easy to manipulate it in order to get as low of a result as you're after. And at the same time, it's just a ratio. So it doesn't actually tell you how much power you are consuming.

Instead, I've been thinking really how to classify it. You are in this data center industry and what you can do. So I started with the colocation level to think "what are the options there?" Because the main difference is that you oftentimes do not own the servers. So you cannot do anything regarding lots of the equipment, but what you can do is you can ensure that you have a supply of renewable energy.

Ideally, you would be producing it yourself. But if not, then at least you're purchasing it from someone. You could try to motivate your clients, be more efficient. Maybe by providing them data on where the majority of energy consumption takes place in real time so that they can adjust it accordingly. Then of course you can try to reuse waste heat and lots of different applications in industrial symbioses.

You could decrease your water use, build your building sustainably because that is something you have a lot of power over, and reuse as much material as possible, not just in the building, but also in the installations that you do have control over. And then if we go further to the enterprise data center, then of course you have much more possibility to green your data center because then of course, you're even running your own servers.

And you're probably writing your own program, so then you could still follow the previous steps and at the same time, you could purchase equipment that can be used longer, especially when it comes to servers that contain so many rare earth materials that have a huge impact when it comes to just their mining and production.

You could refurbish these servers and then reuse them. What is interesting, I've been hearing how the increase in performance between different generations of servers has decreased. So it's possible to refurbish an older server to an almost same performance as the new generation has, and thus avoid purchasing new.

You could try to motivate your IT team to write efficient code.

And something that also has a huge impact is to understand what data is essential and needs to be immediately backed up in case something happens and therefore it needs redundancy. And on the other hand, what data can just wait for a few hours if there is a power outage? And it does not need this redundancy because what we see a lot right now is that data centers are building twice or three times the redundancy, meaning that they have twice or three times the amount of the equipment that they need just in case there is an outage of power or something other happens, and that is very important in certain cases. Let's say if you are a data center behind a hospital. Then of course, you don't want to lose anything for any time, but if you're just storing email or some pictures, then maybe you can wait for an hour or so before you power everything up again.

When it comes to the hyperscalers, what we already see is that they are trying to even build their own equipment. So I think what they can do on top of all of this is to innovate, make even better equipment and then actually reuse their own equipment. Because what is slightly sad, in my experience is that they oftentimes just donate their own equipment or resell it at some secondary markets.

So I think a point of improvement there would be to actually keep it in house, refurbish it, and then use it themselves.

Gael: Well that's very interesting because you mentioned not only energy consumption, obviously, which is a big part of the environmental impact, but you started to mention minerals, resources at large, water. Are these all the ingredients that come into what is your area of expertise, which is a life cycle assessment or is it something a bit different?

Stanislava: Now, I would say that you expressed it quite correctly. It's all of these bits and pieces and ingredients that all come together and although at the moment there's this obsession with the energy use. Of course, energy use is very important, but I think we shouldn't forget everything else that is involved because at some point those things have a large impact as well.

Gael: Okay, thanks. That makes total sense. And Benoît switching to a very operational mode. You recently told me that via Hubblo, you run a full LCA for a CTO, but you are under NDA. So we will not mention neither the company nor him, but could you describe a bit what was the process, more specifically the process involving hosting, infrastructure, et cetera?

Benoit: Oh, the process works. Usually if we address this topic at the company level, it might take into account the workplace as well. So most of the time we have like a consultancy role where we try to assess the impact of it as a whole in the company. So this is very a manual process. This is based on LCA principles, we try to isolate the hotspots of impact. So the huge parts of the impact before we can zoom on specific parts and try to have a more fine grain approach. When we realized that the IT services is the most important part, which is not always the case. Then we could assess those impacts, not only by, let's say human made LCAs, but also with software that could help us to automate the process and make the evaluation easier, repeatable, help the company to be as autonomous as possible and not depending on us to reproduce the evaluation. The idea being that they could evaluate the progress. Are they going to the right direction or not? And so one of our objectives is to become as useless as possible in this process. So it really depends, but for sure if IT services are a big part of the topic for this company, then we could help to automate the inventory of the machines, which is very, very often a pain point.

Because in theory, all companies have a great CMDB that's up to date with all the informations about all the hardware involved in the service. In practice, that's almost never the case except a few companies that are very, very cautious about that, that put a lot of efforts in this area.

Then you have the questions of "what are the impacts of this service?" And Stani mentioned it. When we are in IT, we think a lot about energy. We think a lot about electricity, but that's just a part of the equation. And to illustrate a bit this point based on what Stani said she mentioned like, lifetime of the machines, refresh cycles.

That's something we see often, like companies who say, "It's okay, we can reduce our energy consumption just by buying new servers that are less energy consuming for the same workload as the compared to the previous generation of machines". If you just look at energy consumption at final energy consumption, that approach might work maybe.

It depends but possible. But as soon as you try to evaluate and reduce the greenhouse gas emissions then this is much more complicated. Not to say that most of the time it doesn't work for very simple reasons: most of the time the greenhouse gas emissions due to the manufacturing of the new machines just jeopardize your attempt to reduce your greenhouse gas emissions on the long run because manufacturing has a huge part. And the impact of usage may be not that important, if you look at the manufacturing part. So this is especially true in countries where you are lucky to have a low carbon intensity regarding electricity that you consume. Of course, each phase will be much more important, countries where the carbon electricity is higher. These kind of questions and also how we assess the other impacts, the other operating impacts; and Stani mentioned minerals and metals. That's one of them.

Gael: Going back to the carbon emitted during the manufacturing phase of the equipments, I believe it is called embedded carbon. Where do you get the information?

Benoit: Yeah, there's a whole topic there. For a long time, it has been that you, you had only one choice, which was do you have access to a database where you have impact, factors so constant that you could apply in your calculations to estimate this part of the impact? This is still the case, but I feel like the field is evolving.

Piece by piece, because before you were forced to pay license fees to get those data. That was not the best scenario to democratize impact evaluation and involving companies to take action. So that's one of the topic we worked on in Boavizta. And the first attempt we made was to aggregate all the manufacturing impact data we could get from the manufacturers.

So it takes the form of an open database now. Basically we have some scripts scoring the manufacturer's webpages to get the right PDF files, that's as simple as that. We aggregate the data in the database that you can query. So that's interesting and it gives you some insights about like- let's say- orders of magnitude of the impact of manufacturing a server or a laptop or a screen and so on.

But the thing is that you can't really use that database for evaluation. Because from one project to another, you take 24 inch screen in manufacturer A and 24 inch screen in manufacturer B. The methodologies to evaluate the impact are most of the time not the same.

So it could be almost the same product. You could have different impacts and sometimes the differences are huge. So it's not a good basis, I guess, for evaluation. So we work on on another project, which is an API, where we try to have an approach where you are less dependent on databases.

And how we do that: it's based on scientific papers mostly from Öko-Institut and Fraunhofer Institute in Germany where you can have the impact of one semiconductor, the smallest units in terms of manufacturing IT components. Because it's about what's inside the component, you can then calculate what's the impact of the component, and then if you can calculate the impact of several components, you can calculate the impact of a machine and so on.

That's an open source database as well. That's open source APIs. So how do you calculate? At some point you need data. At least, what is a bit changing now is that you kind of have- it's not perfect yet- but you kind of have access to data without being a consultancy company that has a lot of money to put in acquiring those data.

Gael: But are you telling me that the only source of open source, open data? Actually the only provider of open data when it comes to embedded carbon is now Boa Vista with the IPI.

Benoit: No, I wouldn't say that because I don't necessarily know all the initiatives on the topic. I discussed a bit with people from the CEDaCI project or who I think have a lot of interesting data as well I didn't see how the data is publicized, but that's an example.

And I think there are other projects on the run. So no, the idea is not to say, "hey that's the only way to get open access data and free license". But to say that it's one of the huge topic, at least, we encourage other organizations to provide data.

We encourage manufacturers to open more data. Because in the beginning , that's a bit silly, that you have to build up that kind of project on your own. If you have proper regulation, you would have manufacturers providing data on methodologies that we could understand or that we could verify in some way.

And then we would be super happy and it would be way easier to evaluate the impact of ICT because we would have data from the ground up. So of course it's not the manufacturer itself who does the evaluation. It's a company specialized in that topic. But it's a company that because it is working with the manufacturer has direct access to all the proxy metrics and insights. It needs to make a proper evaluation, which is much harder when you try to do it afterwards. You have to work with aggregated data that you don't know the source.

Gael: Kind of retro engineering, the carbon footprint of an equipment once it has been built.

Benoit: Yeah, that's a bit about that.

Gael: Okay. And that being said, Benoît I have one last question because there are two things that you say that were really music to my ears. The first one being a bit selfish is when you mention that you want to be as useless as possible as soon as possible when you work with a company. And I was like, yeah that's exactly what I say to my clients when I do consulting with them: "I want you to make me redundant as much as possible, though". It's more on the green IT strategy or digital sustainability strategy. But that's something that I was like: you sometimes look at me and say, ";why?", and I'm very happy that you got the same pitch because I really believe that this is what a good consultant should be.

Looking at it, become redundant as soon as possible. But that being said, sorry. It was another topic that I really loved is "the case for automation". And you mentioned that you try to automate measure as much as possible. Could you drop some names or give us an example of how you do that?

Benoit: Yeah. What we try to do with Boavizsta API, you can find it on GitHub is great to evaluate the impact of manufacturing, especially of the servers. But it could be some other context as well. And on top of that, we build several tools. So there is a tool called Bow Agents, which proposes to scan the hardware of a machine.

You ask the right questions to the API and it aggregates the manufacturing impact of this machine as a monitoring metric. So you can get that in your monitoring tools as you use day-to-day basis. I mean, every company running IT services, they have monitoring.

This agent is also connected to another tool which you blow tool this time, but is as always, it's open source Apache to license, which is a scaffold. I think I mentioned this in the beginning of the discussion. It's about measuring the power, energy consumption of servers, the agent, because it's connected to both the API for manufacturing and scaffold for the usage phase it can aggregate the impact numbers of the machine on almost the full life cycle because we still lack good methodologies and data for the end of life, for example. That's something that's missing. On top of that, there is also other project, there's a project in Boavizta called Cloud scanner that will scan your AWS account for all the EC2 instances and give you an evaluation of the impact including manufacturing, so use and manufacturing. This is also based on the API. So that's really an ecosystem. That's a toolbox. And depending on your context, depending if you are on-premise machines, on cloud services or something else. You could select one tool or another, or several of them. We are also working on continuous integration chains for development team. So you can have in your GitLab CI evaluation of impact for the products you are developing from one release to another. You seem to have reduced the potential impact of your application when it'll be in production or if you have made things worse.

So that's an example. That's the kind of thing we have.

Gael: It relates to run the more efficient code that Stani mentioned at the very beginning of the episode. This CICD tool, is it the one related to the SDIA?

Benoit: absolutely. That's the project we are building with the S D I A. Boavizta in SDIA talks a lot to each other because we really like what they do and I think they like what we do. And so we work on this topic together. I know people that Stani knows, but we discovered that thanks to your podcast.

Gael: This is why I love running this podcast. And Stani do you have maybe a success story or just an example of how using this tool has helped developers or I don't know, an agile team or whatever to reduce their footprint?

Stanislava: Unfortunately. I do not really have any success stories because I think it's pretty small scale so far. But definitely there is a growing interest in knowing the impact of computing and doing something about it. It's just that we're at a pretty early stage, I would say. Gael, could I mention something?

I started thinking of this water when Benoît was talking, but then I didn't wanna jump in . So when it comes to the water usage and the WUE water usage effectiveness that Benoît has touched upon. I think one of the problems in the ratio itself is that it doesn't tell you how much water is consumed in total.

And then similarly - and this is also linked to the LCA methodology - is that even if you knew how much water they consume in total. It's very hard to link it to the location, the region where this happens in all these assessments, because maybe using a huge quantity of water in one country would not be such a problem because they have an abundance of water.

While using it somewhere with little water available could be rather critical. So that's another problem that we face.

Gael: Localization is key. Let's save the last part of the discussion to talk about cloud, and I could be a bit the devil advocate here or a bit provocative saying, "yeah, but why don't we move everything into the cloud?" Because if I read Google, Amazon, or Microsoft newspaper, detailing the very same environmental impact that the two of you mentioned during this episode, it seems to be way more efficient to mutualize everything in big public cloud or hybrid cloud or whatever, rather than running data centers on-premise or even colocation data center. Is it something that you agree with or not?

Stanislava: Well, if I can start, I would say I don't really agree with it because at the end of the day, a cloud is not really a cloud in the sky. It's actually located somewhere physically. So you can't just move everything to the cloud because still it needs to be built somewhere and it needs to be operated somewhere.

What I have read from one of these hyperscalers was that they achieve around 90% emission reduction by moving to cloud. But then when you look into that further, it's because they're comparing this with some average data center or low efficiencies and pretty bad environmental impact.

I think it's not really that a cloud is so good, it's just that they have made the data centers that operate this cloud more efficient. You could stay on a physical data center elsewhere, not on the cloud, and just improve your metrics and still have the same impact then as this great cloud.

Benoit: To jump on this one, I kind of agree with Stani and we wrote blog posts on the Boavizta blog, especially on the papers that hyperscalers are showing off regarding the potential of impact reduction when you move your workload from an on-premise data center to the cloud. And yeah, the numbers are calculated on very advantageous scenarios.

But then there are several viables in this equation, the cloud in theory and sometimes in practice, can have good parts good sides:

if you mutualize resources for more services as you mentioned,

thanks to the APIs -and the thing that everything is an API in the cloud-, you can use resources exactly when you need.

That's something that could be done in a non-premise data center, but it's sometimes harder because you need, you need more R&D and more workforce to build tools that make you control the resources as finally as it's made in the cloud. This advocates for the cloud because you could often see on-premise data centers having several tents or hundreds of servers running 24 /7 because they're waiting for the Black Friday.

That's something that you shouldn't have in the cloud because you have all the tools necessary to just consume those resources when you need them, when the traffic goes higher. But that's the technical parts. On the other side, moving the workload from on-premise to the cloud, most of the time it doesn't happen in one night.

So you have kind of a double run. Your service still runs in the on-premise data centers until it's fully satisfying in the cloud. So you have two services sometimes. If it lasts long you could just double your impact. So not really what you were looking for in the first place. But let's say "okay, it's not an issue anymore, we are super effective in moving the workloads in the cloud". Thing is that there is a very different approach from consuming resources in non-premise data centers to consuming them in the cloud. In non-premise data centers, you have order new machines, if you want to deploy any projects. In the cloud, you just have to click or make an API call.

But that's super fast, pretty easy. I've already been in companies where when we looked at the bills at the end of the months, you had clusters of data management that were accounting for 15 K for the months, and it was just a dev environment that was forgotten.

So this hardly happens in a non premise data center because deploying resources is much harder. So you remember about it. That's the full story of FinOps for sure. But in terms of environmental impacts, it has some importance as well. Today, I guess we can do some evaluations of service on the cloud.

That happens. We do it, but we do it on a fixed picture. It's like TODAY the impact of the service is THIS . But it's very hard to estimate and to show projections on: "okay but now that you are in the cloud, your service will inflate, it'll consume more and more resources because people working in your company will have access to those resources very easily, and it'll open the door for many projects".

Some of them would be very useful, but maybe you would also have new projects just because you can do them. So yeah, that's a full question. That's, in my opinion, goes way beyond the only evaluation about impacting the cloud and what's the difference between a cloud provider and let's say a more classic hosting scenery.

Gael: So once again, a multi-criteria approach is needed to answer properly the question. There is no like simple and straightforward answer.

Benoit: Yeah, multi-criteria so that you don't shift impacts from one area to another. That's one thing, but it's also about consequential approach and consequential analysis and not just analyzing the today's picture of the impact. And that's maybe the hardest part. I think, maybe even harder than having a proper multi-criteria approach.

This dynamic view of the impact and the relation between company's activity and what the impact will be in the future.

Stanislava: If I may add I think what is also very important with these studies and impact and multi-criteria assessments and so on is to be very transparent because we have seen lots of new reports or new marketing strategies being published, but they almost never mention what their assumptions are.

And of course the results are gonna be very different if you assume that you are consuming only green energy and you are very efficient. In comparison to if you maybe looked at the slightly more pessimistic scenario.

Gael: Thank you so much, Stani. Because actually, I realized that the way we were discussing transparency issues, both regarding the access to open data, but also the access to what methodology has been used, et cetera. I actually wanted to ask you this question and also thanks a lot for bringing this topic before we close the podcast. And could you maybe provide just one example of an assumption that if you change it, we'll change dramatically the result, something that you've noticed in your life as a researcher.

Stanislava: Well, for instance, just selecting the source of energy would have a very big impact. Even if it was green energy, it could have different impact if it was in different countries. Or for instance, if we look at the colocation data centers, and it's quite popular to be examining the impact of the building.

You know, you just take a flow of concrete in these different databases and if you know nothing about it, there are so many different flows to choose from, which represent different manufacturing practices, different standard of concrete. Maybe you have some that is more durable or not, and the difference in results can be huge.

They can be tenfold, if not even higher. That's why it's very important to be able to rather easily find these main assumptions, how their study was built, just to understand what the results actually mean and if they're applicable to you or not, and if you would reproduce the same study, if you would get the same results or not.

Gael: Tenfold. That was the kind of order of magnitude that I wanted to know and to make sure that I got it right. That huge impact where regarding your methodology and the assumptions underlying it. Okay, so thanks a lot, both of you. You shared already tons of insights, a lot of references.

I believe this is gonna be one of the top five, if not top three episode when it comes to the length of the episode notes and all the references that will be put in it. Still, do you have two last references, thought leaders something that you want to share with the listeners to know a bit better, to understand, a bit better about the data center sustainability issues or potential solutions, or even in a broader way on the sustainability topic.

Stanislava: Maybe in a broader way I would suggest reading, since I really like reading books there are two that I can recommend.

The first one is called The Best of Times, the Worst of Times Futures from Frontiers of Climate Science by Paul Behrens. And what I specifically like about it is that it looks at different topics from both the optimistic and pessimistic perspective and it gives you lots of references.

And the second one . As someone who really likes to understand how things work, there has been a book written by KTH, the University in Stockholm, called Towards the Energy of the Future, and it tries to explain sort of what the challenges are, what is needed in a pretty low level so that anyone can read it and understand.

Benoit: Maybe so not specifically on the cloud but on the impacts of IT and its role, reshaping our societies for a world that stabilizes at 1.5 degrees or less. I mentioned the work of Gauthier Roussilhe, a researcher in France, the report he made on challenging the assumptions of positive impact of ICT on the environment.

Especially there are two reports:

one from GeSI, the other from GSMA. So it's interesting to see that there are very well written papers, peer reviewed papers of amazing quality that nobody knows about. But papers from companies who have a clear and evidence interests in showing one side of the story has echoes on top of the government.

I think it's a real key thing to understand.

Gael: Thanks a lot. That was a lot. I think you might even be a direct challenge to the episode with Chris Adams, where we had, I don't know, 25 references. I think it was a full episode dedicated to, what shall I read? What shall I learn about the digital sustainability topics? But the good news as you mentioned, Benoît, is that we have more and more literature.

The topic is getting traction and hopefully the scientific based papers will get more traction rather than low quality- scientifically speaking communication papers. I want to thank both of you. We covered a lot actually. I think we could have spent another hour deep diving on, you know, codes, green coding, how you do an ACA really like hands-on approach, et cetera.

But that will be a course, rather than a podcast episode. Thanks a lot both of you .That was awesome to have you on the show.

Stanislava: Thank you too.

Benoit: Thank you for having us.

Gael: And that's it. Thank you for listening to Green IO. Make sure to subscribe to the mailing list to stay up to date on your episodes. If you enjoyed this one, feel free to share it on social media or with any friends or colleagues who could benefit from it. As a nonprofit podcast, we rely on you to spread the word.

Last, but not the least. If you know someone who would make a great guest, please send them my way so that we can make our digital word greener one byte at a time.

❤️ Never miss an episode! Hit the subscribe button on the player above and follow us the way you like.

📧 Our Green IO monthly newsletter is also a good way to be notified, as well as getting carefully curated news on digital sustainability packed with exclusive Green IO contents.

Uncanny Magazine Podcast #58A

Uncanny Magazine Podcast

Exciting News, Patreon Support, and Story Introduction

The AI-powered Podcast Player