Rudolf Meier (SAP) - Carbon Calculation Basics

This video belongs to the openHPI course clean-IT: Towards Sustainable Digital Technologies. Do you want to see more?

Enroll yourself for free

Rudolf Meier (SAP) - Carbon Calculation Basics

Time effort: approx. 46 minutes

An error occurred while loading the video player, or it takes a long time to initialize. You can try clearing your browser cache. Please try again later and contact the helpdesk if the problem persists.

Scroll to current position

00:03Hello everyone, welcome back to our Clean-IT Open Exchange meeting.
00:07My name is Nils König and I'm as always here today with Sophie
00:11and Max from the HPI Student's Sustainability Club and the Clean IT.
00:17Today's guest is Rudolf Meier. Rudolf Meier is the head
00:20of performance and scalability in product engineering at SAP.
00:24He joined SAP in two thousand and four and has been engaged
00:28in performance scalability, benchmarking and sustainability
00:32topics there ever since.
00:34Also he holds an experimental physics postgraduate degree from
00:37the university of Tübingen.
00:39In the clean IT initiative, we often talk about the carbon
00:42footprint of digitalization and software, and especially how
00:47it has been rapidly growing over the past decades.
00:52However in order to fully understand this problem, and also
00:55to find and compare different solutions, we first need a way to
00:59calculate the carbon footprint of, for example operating software.
01:03And that is exactly what we're going to talk about in today's
01:06talk on Carbon Calculation Basics.
01:09Welcome Rudolf Meier.
01:13Thanks a lot Nils for the introduction. It's very nice to be with you here
01:19for this open exchange. I hope you are all doing well
01:23and I will push it off with a recap of my Clean IT contribution
01:28on carbon calculation basics. But before I start into that, let me
01:32give a bit of context. In this time of
01:36climate action, we all have to look for how to make our life more sustainable
01:42and a digital technologies that's one way to do it. Our digital
01:46technologies have already started to contribute and the example
01:52which is on everybody's mind is just in front of us, that's a communication technologies.
01:57So during the pandemic,
02:01work went on in home office and
02:05communication technologies enabled us to do that, and at the same time
02:10the commute and travel emissions significantly went down. So
02:16here we have some contribution. It's maybe not always to our
02:20liking so it would be also nicer for me maybe to sit in
02:25together with the hosts in one room, but I think that's a
02:29common thing that we will see that not everything which
02:34we have to do to save the planet will be to our liking.
02:38Ok so now technology teaches technologies can of course do much more
02:43and me coming from SAP, of course one of the things which we
02:46can do is optimize processes. Now that's starting from optimizing manufacturing,
02:56transportation management or the circular economy.
03:02There are lots of possibilities. There's also the topic of transparency. So
03:08software digital technologies can give us transparency on the carbon footprint
03:14organizations and enterprises, of course in the real world and having
03:20transparency of course gives an opportunity to compare, to have competition and ultimately
03:28get optimization in that direction.
03:31But of course we have to make sure that the underlying software systems
03:36are designed in a way that we
03:40do this efficiently, that we use our resources in a way that is
03:47good for the environment, that we need to use a minimum of resources for that.
03:52And this in particular of course calls for sustainable programming.
03:57But before we go there let's first try to get a basic understanding
04:02how running software connects to energy consumption and carbon
04:08emission. So let me start. So I should share my screen.
04:15See I get this right?
04:23Can you see my screen? Yes.
04:26Ok perfect. Then let's go into a little bit of
04:32let me get rid of this, a fun warm up. Now that somewhat
04:37what I want to do is a Back-of-an-envelope calculation, how
04:41do we get from compute to energy and co two just to kick it off.
04:46Now and the question we want to ask and answer is, what is one
04:49CPU second in electric energy consumption in kilowatt hours,
04:54and emissions in kilogram of co two. I like these back of an
04:58envelope calculations as
05:01originally I'm an experimental physicist, that we do that all the time.
05:06It's very nice to get the lay of the land, but please be warned
05:11there will be lots of assumptions or quite some assumptions.
05:14So don't expect an exact result. It will be more like an order
05:18of magnitude which we get for that.
05:21So what do we do. First we get some idea of what one kilowatt
05:24hour is. That's really easy. These days with all the electric
05:27cars because this is something which moves your electric car by five kilometers.
05:33So what's the carbon emission connected with that. That's a
05:37bit more complicated because of it depends on how you generate your electric energy.
05:44For simplicity, I here just assume the world energy mix. Of course,
05:48this can be very different. Either if you catch with the wind
05:52or if you burn coal will be different but the world energy mix is about
05:57half a kilogram of co two per kilowatt hour.
06:01So we have some idea of how to calculate that. Now to get to
06:05the CPU second, we have to pick a typical our server CPU, so
06:10I'll stick with server CPU and here I picked an example. This is
06:16picked the right way so that the numbers come out easy, so we have a
06:20realistic what we have here for our present day
06:24hardware. We have hundred fifty watts for a CPU with forty two threads
06:30which would give us three dot six
06:33watts per thread and then ok to get to the energy, we say
06:38oh we use this thread for one hour, that would be three dot six watt hours.
06:44Now and this would be three dot six thousands of a kilowatt hour in energy.
06:50And then we want to go for the CPU second, so we just have to divide
06:55this amount of energy for one hour by three thousand six hundred.
06:59So we end up at the very nice little number of one
07:04millionth of a kilowatt hour per second.
07:07And of course with the assumption appear we can translate that
07:11to the corresponding carbon emission.
07:14Would be half a millionth of a kilogram, so
07:19of co two and we are arriving at all rule of thumb CPU second
07:25corresponds to one millionth at our in energy and half a milligram of co two.
07:32Just to make this a little bit more tangible with the co two
07:35water half a milligram of co two if you run this thread for
07:39an hour, we have produced a liter of co two approximately. That's
07:44what you have for that.
07:46Now you can play similar games with other resources, for example with memory
07:51and then you can end up with one gigabyte of memory
07:55using use one gigabyte of memory for one second corresponds two
08:00point one miliwatt hours in energy and of course the watts of the equivalent
08:06amount of co two produced.
08:10Now of course with numbers like that, one can immediately go
08:14in the direction of what to do for sustainable programming
08:17and what jumps into the eye immediately here is that one could
08:21imagine yeah let's save one thread of CPU
08:25by just using one gigabyte of memory in addition,
08:29and that would with these numbers
08:33have a factor of ten energy saving connected with it, and that's
08:36of course also the metric which is behind caching, buffering, indexing
08:42which we can have in sustainable programming to support us in
08:47saving resources. But today we are not going to go deeper in
08:53sustainable programming unless you want to go that direction.
08:57What I want to do is go from the microscopic here from the
09:00CPU second what does it mean to operating software in data center.
09:06That's where I come to
09:09a recap of the slides which are used for my contribution to the Clean IT
09:15forum and there we start with just setting
09:20the stage. Yeah, what are we going to look at, and that's
09:26what are the contributions to carbon footprint.
09:29I just mentioned a few on the side it's all symbolic because
09:33depends a lot on how you produce your energy or generate
09:38your energy in each case but what you have here as
09:43you have manufacturing which comes in transportation, comes in
09:49operation, comes in that's the one which we are going to look at and others.
09:54If you want to go deep here, this is an own science. Just a
10:00search for greenhouse gas protocol for this and you will find
10:06all the possible contributions which come in here. What I want
10:09to focus on is what we directly see in operation. So this
10:16operations the slice
10:20and what is contained in that operations the slice that is the
10:25consumption of IT equipment in the first place at the direct consumption.
10:29So we have compute storage service resources, we have CPU memory disk
10:36network which contributes to spending
10:41electric energy, and of course there's indirect consumption
10:45because the data center doesn't just have this inter infrastructure,
10:49but also cooling, lighting and other things.
10:52Now in my contribution I had a four-step
10:57approach here. The first two steps are actually small steps
11:01because they just introduce important quantities, we'll go over that quickly
11:07and then the third step that's the one where I was focusing
11:11on most how do we get from business requirements that someone
11:16knows what to do in in a company, what processes to run to,
11:21a prediction for the carbon footprint of the hardware which is running
11:25behind that. And in the fourth step we see how this order evolves into the cloud.
11:31Ok let me just do these first two steps
11:34briefly. For the first step, we assume that we have exclusive data center usage
11:40so we know how much data center electric energy we spent for, or software
11:47that is just measured, or
11:50we will know that and then how do we get to the
11:53carbon emissions that's a by the way of the carbon emission intensity,
12:00carbon emission intensity
12:03to be multiplied with the total energy which is measured
12:07and gives us the carbon footprint.
12:10I already gave a number for that in my warm up but normally
12:14this is something which you will
12:17get provided from your electricity provider, and here again the
12:22global average four hundred seventy five grams per kilowatt hours,
12:27and just a side note in SAP data centres we have
12:31zero grams per kilowatt hour by use of hundred percent renewable electricity.
12:37Okay so first quantity which is important
12:40CEI and second one now we make it a little bit more realistic
12:46we don't have the complete data center available for
12:50our software test, we just have dedicated hardware, and the question
12:54is how do we get from the electric energy consumption of the
12:58dedicated hardware. Eraw to the carbon footprint and that's
13:04a two step process now. Our first we have to use a factor which is
13:09called power usage effectiveness at the
13:14number which is specific to the data center you're running and
13:18you multiply your Eraw energy get the total energy and then
13:22the next step of courses as the same as before
13:26to get the carbon footprint. This and multiply it by the carbon
13:30emission intensity of your electricity provider.
13:35Okay this PUE is a very interesting quantity because it describes
13:39how efficiently a data center uses energy just
13:45as the ratio of total facility energy over IT equipment energy and
13:51maybe this is a bit too short. There is a standard on how to calculate it.
13:56So in the details of course that's not as easy as I put it
14:00here. One has to do something but it's well
14:03defined of what we have here. What's a typical number here.
14:09Typical values are around one dot five I write here. But there
14:13are better values also. There are big data centres
14:17which publish numbers down to one dot two or one dot one. Tbere's
14:22a lot of effort going into getting this
14:28additional energy down, which is beyond the pure IT
14:34equipment energy.
14:37Okay so that was just the preliminary steps
14:42which we do and now we come to the main step and the main step is
14:47how do we get from business requirements to carbon footprint?
14:53Now for this we use something which we have in place for a long time at SAP
14:59and that is simply if you know your business requirements, a
15:03customer comes in the traditional on-premise business, a customer
15:07comes and asks what's the hardware I have to buy for that. So
15:12that's the first step- translating business to hardware requirements
15:16and here we have a sizing procedure which is available here so business requirements
15:23can be a lot of things. There you could ask how many sales orders
15:27do I want to
15:30process, what do I want to do in logistics, in manufacturing.
15:35So there are lots of processes which could be involved here and
15:39if I know what the the sizes are, the throughput which I expect
15:45and if I also can specify a profile, yeah so I will have a peak hour
15:50of so and so many
15:53users running in parallel with that kind of
15:57a business process.
16:00Then I can put this into a questionnaire, which is provided by
16:06the quick sizer. So the link to the quick sizer I think in the
16:10original contribution
16:13and this returns some
16:16numbers for the hardware requirements. So memory and disk we understand immediately.
16:23Saps you might not understand immediately. This is a specific
16:27compute power
16:30evaluation number which is provided from
16:35SAP through the SAP benchmarks. So there's the SAP benchmark which
16:40is run by hardware partners,
16:43and they have then afterwards a number which specifies for usage with
16:49SAP software how strong is that
16:53hardware in terms of compute. So there's a lots of benchmarks
16:58being run by all different hardware partners also by by hyper scalers.
17:04And then one can from this define
17:09the amount of or that the typical hardware which is needed
17:14for the result which we get from the quick sizer. So the quick size specifies
17:18hundred thousand SAPs then we go here and check what hardware
17:22has a hundred thousand saps and that's the hardware selection and
17:27then we are ready with
17:29the hardware requirements. So I just talked about computer but of course memory
17:34disk network also are going into the hardware selection.
17:40So if you have the hardware selection ready, then you can also
17:46predict what this or estimate what this should
17:50be in the
17:53carbon footprint. So first you go to the hardware partners, you
17:57get the power specification for the hardware, from
18:01the power specification will specify what's the load you will
18:05get, ideal you will get a better at high load or at maximum load and then you
18:15put in the load profile which you will get out also as a sizing result.
18:21You have some hardware load power relation
18:24which of course depends on the specific hardware also on the mix of
18:29cpu and memory which you have to run and then
18:32you finally you go here to the
18:37final calculation where
18:40with the energy of the raw energy which you get out of this calculation, you go
18:44and multiply it by the two factors which we mentioned at the beginning,
18:47and you end up in your carbon footprint.
18:51So now in the contribution of course there is somewhat details
18:56on how we do the sizing and where to find
18:59or the sizing resources which we have at SAP for business
19:05there, and then we also had
19:09an example. I think that was an example where
19:13we just said, okay we get out of the quick sizer, we get a midsized server
19:20with four socket and three terabyte was
19:24approximately what we
19:25got out. So that has a certain capability and running the business
19:30and then we see what the full power load is. We check what the
19:35average load is at the
19:40load profile and then we get out a number for the full year
19:45for the operation here,
19:47including the PUE and then of course using
19:52a CEI that would be six dot two tons of co two.
19:56Now of course this is, you can say this is easy. We own the hardware and we
20:02put it on there, we can also afterwards measure now that's here we predict
20:08the energy which will be taken and
20:13what we can also in principle measure if he owned this so for on premise
20:19that's a very simple procedure to do.
20:23If you look at the details, it's maybe not always
20:25so simple, but the path is
20:31is right so it's a question what changes if we go into the cloud.
20:36Actually in terms of the calculation for the
20:41owner of the cloud, is not so much that changes. It's just that
20:45there are some things which improve quite a bit if you go into the cloud,
20:50and that's the two things which I want to mention for the cloud.
20:55First we have the advantage of scale there.
20:58So not everybody is bringing their own resources, their own hardware
21:03with so we have the possibility of sharing the resources. So
21:10we can get a much higher, more uniform utilisation are in the systems.
21:15And that results in a net energy saving for
21:19the service consumers. And of course if you go into the cloud
21:24that and you have also the advantage of choice
21:28with respect to a carbon footprint. So for example in the SAP
21:33cloud to the carbon footprint is zero by use of hundred percent renewable electricity.
21:41Okay so this I wanted to kick it off, and
21:45please no take over again
21:49moderators and then we will go into the Q and A.
21:54Yeah thank you very much for your presentation.
21:57We will now proceed as already said with the Q and A session as always. So
22:01if you have any questions you can either raise your hand and ask your question directly
22:06or you can also put it in the chat. I have seen we've already
22:09had one question in the chat.
22:11Just as a quick note, we are still recording
22:13so we can upload the video to the Clean IT forum afterwards.
22:16So if you don't want to be recognized, you can of course just
22:19keep your camera off when you ask a question.
22:23I just want to ask a quick first question for clarification. So
22:28this PUE score or this index of
22:32this has to be effectively calculated for each data center
22:35in the world uniquely, right? So this is not anything that can be
22:40measured across data centers. It's just for each individually.
22:44Yeah, it's the operation of each data center of course depends or
22:49that this influences a lot what the outcome here is. So there are
22:54better and better techniques to run this. So the best ones in
22:58the field I think, they run at a high temperature. So they
23:02have twenty seven degrees or so in their data center. They cool
23:07with water or with outside air, they place their data centers in areas
23:12which are cold anyway, and if you look at their reporting, you
23:18see a clear dependence on the seasons for the PUE, because
23:23cooling is of course a big driver.
23:26And otherwise but it depends on the data centers and
23:31it's standardized how to deduct from these numbers, but it will
23:37be very dependent on how you run it. So there's no way from the outside
23:41to determine that unless you maybe know I think there's no way.
23:46Even if you other that the power supplier
23:50you don't know how much IT equipment is going to be powered by this.
23:55Thank you. Okay, so the first question of the chat,
23:58and I think that was asked during your when you mentioned that
24:03you calculate zero carbon emissions for the SAP clouds was
24:07because of the renewable energy,
24:10and the question is do you also include embodied emissions
24:15into this calculation?
24:17This is a which, which emissions? So we
24:22it is I can actually, I can
24:24ask the question directly. So basically,
24:28what you want to basically do is you calculate operating emissions,
24:33when you say more or less that the data center uses one hundred percent
24:38renewable energy, it's zero but in practical terms, it is not
24:41because you have to manufacturer and the hardware, you have to.
24:46So if I said renewable energy, then it was wrong. It's renewable electricity.
24:51So it's the electricity. This is a good point, so
24:57for electricity we have been
25:01using renewable electricity, of course in part by just having
25:07a power generation. Also SAP with
25:11a PV and with
25:17renewable energies, but of course also by compensating with the appropriate and
25:23high quality certificates. That's how you would do it if you
25:27buy your electricity from your energy provider and ask for the renewable energy.
25:33Ok, so that we have had for since two thousand and fourteen
25:38and now what you ask is for the complete picture, yeah?
25:42It's not just electricity but we have manufacturing. We have
25:47the buildings, we have heating and things like that and here
25:52we are not at zero yet, but we have committed to have, be
25:56net carbon neutral by twenty twenty three.
26:03In your calculation, this does not show up in your calculation. Is that correct?
26:09Well as the embodied emission when you at the end you also purchase a service
26:15or you purchase storage arrays and everything and then these
26:18storage arrays they have to be produced
26:22and of course also has as it has emission factors,
26:27and then after the life cycle you basically have to recycle
26:32it in a certain form and that has a consequence, are you taking
26:37this into calculation or into into consideration when you do your calculations.
26:43Well I haven't seen it and I was just.
26:46I try to say this in the very first slide
26:49that we just look at operation for what I'm doing here.
26:54Okay then it's clear.
26:56So this is why, this is why we have this slide to make clear
27:00we are looking at operation at electricity use in data centres.
27:05That's what the focus is here.
27:11Okay so I think this question has been answered.
27:15The second question from the chat is from Vincent.
27:18In the presentation you only focus on what is happening at
27:21the server-side, but when you deliver software in the cloud
27:24you also need to take the network infrastructure
27:27and the web into account. And the consumption of the terminals as well, right?
27:32Yes, it's a question on what what is really in
27:39the scope of your calculation. So what I was focusing is on
27:43what is happening in the data center, but if you of course go
27:48downstream, yeah so that would be what is also the consequence
27:53for using the software. Of course then there is also
27:58energy usage on the front ends. It's
28:04no question that this is something which
28:07also will come into play in the complete
28:12picture for carbon usage. So there is of course
28:16it's a question when do I do a double counting, where were to I
28:22attribute the different contributions to. So there are
28:26in the moment there are processes being defined how you attribute that
28:31to your supplier, how you attributed to energy used in your own environment,
28:37and how you attribute it
28:40further down the supply chain. So that's a
28:44very nice reading as the greenhouse gas
28:48protocol for scope three emissions. So that I think there's
28:53no question that if you have the complete picture then
28:57you also have to look downstream and upstream but you have
29:00to have clear rules where where are the boundaries, and where
29:04do we start doing double counting.
29:08Yeah I hope that also answers the question.
29:11I also have a question that kind of relates to this.
29:15So you've already mentioned a kind of the boundaries of the
29:18calculation that that you are doing or the methods that you are using.
29:22So want to ask, do this
29:25method survey improved so are you currently working on maybe improve
29:30including the embodied greenhouse gases or maybe including
29:33also these upstream and downstream carbon emissions?
29:37As I said SAP has pledged to be
29:42carbon neutral by twenty twenty three and by then we have to include everything
29:47that's clear. Okay.
29:51But do you also have developed like a more sophisticated
29:55method so that other companies could maybe also include.
30:00So for that I have to say what I did here that's the basic. So it's not
30:05the most sophisticated ways of how we calculate
30:11carbon footprint for a company. For that we have
30:15products. We have product from a suite of products in
30:20its called climate twenty one. And there you have, for example
30:24coming up a watch tower for GRI, so that's the Global Reporting Initiative which gives
30:31an overview over all the
30:34possible angles of sustainability. That not only includes the
30:40climate action but also other things, like a circular economy,
30:44waste management, also diversity and that
30:50allows companies also due to get an overview over
30:56the carbon emissions which come from their supply chain.
31:00So in particular for manufacturing you can look at the carbon impact
31:05of different goods you are using, you can compare different
31:10providers, suppliers for for these goods and then optimize your business
31:16accordingly. So it's
31:19for talking about what can a company do to optimize their we have a lot of
31:26work going on and what I would recommend is we just that last
31:31week we had the Teched- SAP Teched and there is a very
31:34nice talk or a demo also on the possibilities you have there.
31:40You can find it on Youtube. So we have forty eight hours, I think
31:43of a Teched material on Youtube now and if you look for sustainability
31:49and SAP and Teched, you should find it.
31:53Okay thank you. I think we can of course include this link to the Youtube series
31:58maybe in the recording in the Clean IT forum
32:01so everyone can look that up. I see Maxi raising a hand.
32:06Yeah I have a question to you. You told that SAP is providing
32:10a questionnaire to calculate the hardware requirements.
32:14When can coming to SAP cloud, are these clients renting the
32:19dedicated hardware on SAP cloud are is that weight of optimizing it?
32:24Yeah so this questionnaire, so what I showed is
32:30clearly for an on-premise customer who wants to know what hardware to use.
32:35It's very different in the cloud because the customer doesn't really care.
32:39In some cases, there are some match points of course. Because
32:43if you go from from a on-premise solution to a cloud solution,
32:49then you know some things already how much footprint you have, and
32:55there in principle what you want to know is more
32:59like what kind of service plan do I have to
33:03buy and there
33:07that the sizing and also the quick sizer has a one possibility
33:12for the cloud. So you can run the cloud quick sizer and what
33:17you get there is information on what kind of service you want to buy
33:22for your
33:24profile of business processes which you put into the questionnaire.
33:29So that will be quite different, right of course. It matches
33:32at some point. It has to match in resources but that's not on the customer side
33:37of course. There it's a more the question if you have such throughput
33:43in a certain business process and selling
33:47and sales order and distribution, then you get
33:53the result you should buy
33:56an m-size plan for the service which is connected with that
34:01for x four hundred cloud, for example.
34:03Ok and on SAP cloud SAP is optimizing like auto scaling.
34:09Exactly this is the one which I mentioned on the last slide.
34:13There you have really big possibilities to
34:18make sure that you are elastic, that you are scaling, that you share resources.
34:26Of course there are challenges which come with that for for the provider
34:30but the customer doesn't have to
34:33care about this. So if you share resources, you also
34:36have to make sure that the resources are shared according to
34:40what was bought by the customers, things like that. But that's
34:42I think it's a no brainer that you have to do that, but
34:46what you can do is of course use old features which you have in
34:51virtualized environments, which you have with
34:55also down and up and downsizing
34:59the resources, which we give for a certain customer depending on their
35:05profile. So there are also
35:09actually we also have a product which is called a LaMa Cloud
35:16which is offering a service which is not directly a cloud
35:20service but it offers a service that SAP instances which are
35:25maintained, or which are run
35:28at hyper scale as they
35:31should be run effectively. And the first of course what comes in mind
35:35is of course always of the cost. How can I run my SAP instances
35:40cost-effective and there you can for example build profiles
35:44to see how you can scale up and down these SAP instances
35:49to make the most out of the purchased infrastructure
35:54at the hyper scaler. So that's
35:57possible and of course
35:59that's maybe one point
36:02which is a maybe hidden point, but is itself
36:06pretty clear that very often price and resources go parallel with carbon footprint.
36:14So and of course I'm the first concern of
36:18people at least of the financial part of companies
36:23is always the price. But that
36:26is a very nice synergy because if you need less resources and
36:31hence you pay less then you
36:34normally, you get also a better carbon footprint out of this. Thank you.
36:43Ok so we have two more questions I think. One short
36:47question in the chat is-
36:49do we have an idea or maybe do you have an idea if the consumption
36:52of data centers is a significant part of the overall carbon footprint?
36:56And I think the overall carbon footprint of the
36:59digitalisation as met.
37:02Yeah I think about that is of course that's a big debate because there is
37:08if you look at their some contributions may be about that the introduction by
37:12by professor Meinel already mentions this that there is a rising
37:17part of digital technologies in the overall picture. If you have a few
37:26typical examples where there is a lot of energy consumption and
37:32a lot of part of the total energy consumption going into digital technologies and
37:38maybe the other key words here would be
37:43bitcoin and would be streaming. So that's the ones which are always in
37:49the headlines for really generating a lot of
37:55energy consumption and carbon footprint.
38:00But I would say that with digitalisation. We of course also have an overall
38:05increase in energy consumption there. What is always the question is,
38:11is this going to be
38:14making the situation worse or is it going to make the situation better.
38:20So with something like bitcoin you can of course argue that
38:24this is not going to help at all. Also there is there's alternatives
38:28also. There are also other
38:31currencies which are doing better in that way. But here you could,
38:36people have argued that this is really a bad way of wasting energy.
38:41But then you can also look at other
38:45applications like zoom.
38:47With Zoom, as you can say of course with the big explosion of
38:51zooming and in the world the last two years
38:55of course Zoom is going to take more energy in data centres. But of course
39:00it's bit hard but it's
39:04close to everybody's mind what happened.
39:07What happens if we meet here in Zoom and not all in person at HPI.
39:13Now so I think it's pretty clear that there is some impact
39:17and even for block-chain, one can argue we applaud changes connected with
39:22bitcoin. Even there one can argue that there are examples where a block chain
39:27can save energy by for example you use block chain to make sure
39:33that a certain product you have bought is really up to the standards
39:38in particular to green standards
39:42and you recognise counterfeits or you to prevent people
39:47from buying counterfeits and essentially
39:51preventing them from buying the CO2 heavy products.
39:57So there you can argue even with an expensive technology
40:01that can also be on the positive side of energy saving.
40:06One just has to make sure that the overall is going down.
40:11And so I think, for speaking for what we normally do process
40:17optimization and companies, this is normally always going to have a
40:23big positive impact, like a standard example transportation management.
40:28If you do transportation management right you just save a lot of kilometers driven
40:34and a lot of planes flown and containers being transported
40:39by doing it the right way. So
40:42I think overall, I think it's rising -the data center
40:48energy consumption. It has different reasons why it rises.
40:54There are reasons which are good reasons and there are reasons which we should
40:58maybe ask if this is really necessary.
41:02Yeah I think you have mentioned the very point of the
41:05entire clean IT initiative. So actually
41:08clean or digitalization can help at many points, but
41:12we always have to make sure that the tools we use are efficient
41:16themselves or efficiently designed.
41:18There's one last short question to clouds in the chat. So
41:23assuming for the case of SAP cloud services, there are probably
41:27millions of end users with their devices powered by electricity themselves.
41:31How or will this even add to the carbon footprint calculation or other
41:35calculations where this included?
41:38It will certainly be included in calculations according to
41:44greenhouse gas protocol scope three. There it will certainly be included.
41:49The question what will be, how will we distribute that between the
41:57consumer and the supplier and that is something which will
42:02be something which the future will show because there are lots of
42:06also standardizations and reporting standards being discussed
42:11in the EU and in the UN
42:15to see how this will be
42:17reported. So what is of course the over arching idea is to reduce
42:23the emissions in total. So it's important to have transparency
42:28and I think you're completely right or whoever asked that question
42:32completely right that one shouldn't forget that part.
42:35It's then the question where does it show up in the
42:40overall footprint and how do we get it down.
42:46Okay then I've just one last quick question from myself again.
42:51And I want to ask how widespread is the use of these carbon
42:55calculation methods. So you've mentioned a few products already from SAP which
42:59use these calculations or which offer these calculations,
43:03but how widespread is this method in other companies may be and
43:07what does it mean that there are being used thereof?
43:11Okay so it maybe I should not talk about other companies too much. So
43:16that I can do a general comment. The general comment is that
43:19it's really rising. So you see
43:22carbon calculators popping up in various areas. So I think it's
43:28pretty clear that this is something we need and
43:31with all the actions of climate actions and with
43:35regulation also, that will be very important.
43:39So and I would suggest just searching for carbon calculation
43:46on the web and you will find your answer.
43:49I could name names now but this is no point there.
43:53Okay thank you very much. I think I see no more questions
43:57or raised hands anywhere, so
44:00I think that's it for today. Thank you very much for presentation
44:03and for answering all of these questions.
44:06Just for the end I will quickly share my screen again.
44:09So I hope you can see this now.
44:13The next live talk I just want to quickly mention that I will
44:15be at the thirteenth of December at twelve.
44:19The thirteenth of December. So it will be in three weeks and not four because
44:23four weeks would be very close to the Christmas holidays already.
44:28And the next live talk will also be a little bit later so this
44:31will be on sixteen o'clock p m after the Central European Standard time.
44:35So four hours later than usually. And our next guest will be Mathias Renner
44:41which who will be talking about sustainable software engineering
44:44which I think you've already mentioned today a quickly
44:48and the opportunities and barriers of these.
44:51So for the end just if you haven't already but I think most of
44:56the people here in the Zoom room have already joined the Clean IT Forum.
45:00And as always if you have any questions or any feedback or
45:03anything else you can always contact us by a Clean-IT@hpi.de.
45:09So with that being said, thank you again very much. Thank you for your questions
45:13and for attending today. Thank you for presentation and for all the
45:17very kind answers and I hope we can see you next time.
45:23Thanks a lot.
45:25Goodbye, everyone. Goodbye.

About this video

Software can play an important role in reaching sustainability targets. It can enable efficient business processes, or provide carbon footprint transparency. But of course, running software itself requires electrical energy. In this clean-IT openXchange live talk, Rudolf Meier presented basic concepts for estimating the carbon footprint of operating software. After the presentation, the participants also discussed further about embodied emissions and how emissions from cloud usage and infrastrcuture could be included.

Rudolf Meier is head of Performance & Scalability in Product Engineering at SAP. He joined SAP in 2004 and has been engaged in performance, scalability, sizing, benchmarking, performance data mining, know-how transfer, and sustainability topics ever since. He holds an experimental physics post-doctorate degree from Tübingen University.