Speaker 1: 0:22

Welcome to a new episode of TechSign Talks on Tour. We're at HPE, discover Barcelona. We're here with Pascal Lecoq. He's Worldwide Director of Sustainable Data Center Modernization at HPE. Good morning. Good morning, that's a long title Worldwide Director of Sustainable Data Center Modernization. So you help your customers make their data center more sustainable, I guess.

Speaker 2: 0:47

Yeah, of course, of course and definitely. In fact, our motto is also from end to end, meaning from the cheap to the cheer. So I love this kind of short sentence because this really represents what we want to do with our customers helping them in their whole transformation, in their whole IT transformation, them in their whole transformation, in their whole IT transformation, meaning that typically, I mean in my service organization, what we do this is to help, from the IT to the facilities, to transform the IT. So we are not focusing on application layers, definitely, but I would say on the hardware part of it. So hardware for the IT, you know, and hardware for the facilities, for the data center. So at this stage, we take the whole, let's say, spectrum of all the activities related to this transformation in order to make it, of course, best in class with the best practices in terms of technology, but also reaching their sustainability objective globally, from IT to facilities.

Speaker 1: 1:47

Okay. But if we look at data center modernization, there have been some studies and researches, for example from Gartner, and they usually tend to point towards the public cloud as being the most sustainable. I'm guessing you probably don't agree with that, but they focus on the flexibility of scaling and the mass of a public cloud, that it's easier to be more sustainable because those data centers are usually carbon neutral, etc. Etc. But you help your customers with an on-prem data center become more sustainable. Can you compete with those public cloud or is it? Can you enlighten the listeners in how that works and where we're all at at this moment?

Speaker 2: 2:35

Okay. So let's say, in the great majority, the public cloud. You know, those hyperscalers, definitely what they do in terms of sustainability, this is, of course, to have data center design that are, let's say, coping with sustainability objective in the sense that they use now more and more and massively green energy to fit the board. Then, in terms, I would say, of energy efficiency of the data center, the fact that there are energy efficient and this is true does not really fit in the sense that the design is perfect. You know, because what they do, they got a standard design, you know, and then they apply. That you know, they copy paste everywhere in the world.

Speaker 2: 3:21

It's not one of the most optimized, I can guarantee you. But the thing is that in the public cloud they rely on the fact that they reach much more utilization ratio of the IT than in an on-prem data center. In an on-prem data center, the average utilization ratio something like 30 to 40% of the capacity of the data center that is installed. So of course, the PoE suffers because of this low utilization ratio. The fact is that through consolidation, massive usage of the IT among all the various stakeholders, clients, customers that are using those public cloud they managed to optimize the utilization ratio of the IT, the data center, that is more something like 70-80% average, meaning that they can reach much better PUE due to this fact.

Speaker 1: 4:16

But with HPE GreenLake you also lift the utilization right, because you offer the same scalability.

Speaker 2: 4:22

This is what we are doing, so through private cloud or hybrid cloud, because definitely our strategy is hybrid cloud and we help our customers with that, because we got a GreenLake colocation activity also so that the environment that are much more used to reach better energy efficiency and so better sustainability.

Speaker 1: 4:54

And with GreenLake colocation you also have a bit of control about the energy being used.

Speaker 2: 4:58

Absolutely.

Speaker 1: 5:00

Okay, but is it a fair comparison? Because I guess your customers make the comparison between public and on-prem. But can you come close to their utilization and emission ratio?

Speaker 2: 5:16

They make comparison, but only criteria to decide about the strategy. So where to host your environment. Will it be on-prem, Will it be off-prem, at a colo, so including with GreenLake, or will it be on-prem? Will it be off-prem at a colo, you know, so including with green lake, or will it be in the public cloud? Definitely, I mean the cost and I would say the carbon footprint is not the only criteria. You know you get criteria like guarantee of performance in terms of sla, but also in terms of security, in terms also of sovereignty of the data, to be sure that the data is, for example, stored for critical application, you know, in the country or in the EU or etc. And not subject to cloud act or patch real tact, you know from the US. So there are a lot of criteria like that also that are part of the decision from the customer.

Speaker 1: 6:05

Okay. And if we move into the data center, I'm noticing that HPE is becoming more and more also a software company, besides all the hardware they offer. So I wonder how much influence the software have on sustainability versus the hardware.

Speaker 2: 6:22

This is definitely very important. So, typically, when we start working with our customers on their transformation journey, we usually have, let's say, an activity that we call advisory, consulting, kind of strategy consulting on what to do, how to do it step by step, what is the roadmap to reach the target, et cetera. And in this area we got several pillars and typically three pillars that we, let's say, analyze with the customer. So one is the IT efficiency in terms of sustainability. The other one is, of course, an obvious data center efficiency in terms of sustainability, but also software layout in terms of sustainability. And why?

Speaker 2: 7:07

Because, depending on the use case, depending on the business that the customer runs, the fact that you will, let's say, develop your application or migrate your application with a software toolkit A or software toolkit B can have a real impact on the IT usage and so on. The power that you will consume. This can be around from 1 to 4, typically. I mean if this is developed with the right tools or if this is not developed with the right tools. And so if you get a ratio from 1 to 4, this is the same impact on the IT resources that you use. You will consume four times more CPU memory, etc. So you will consume the same kind of ratio also in terms of electricity you offer consulting in this, or how does it work?

Speaker 2: 8:00

absolutely so. We. We have dedicated tools that we have developed from HPE. This is called WriteMix Advisor. That is able, from an existing stage, we can take an assessment of what is currently running, model it in our WriteMix Advisor, rma, including sustainability elements, because RMA is also capturing, let's say, all the configuration, physical configuration, to model and simulate the carbon footprint typically, and then we can do what-if scenario. So if we move, if we consolidate, if we virtualize just to talk about physical, virtualize, control a date, those kind of environment on new IT equipment or keeping the same equipment, but through, let's say, consolidation, if we move to the cloud, to a private cloud, to a public cloud, etc. What will be the target in terms of IT environment, but also in terms of carbon footprint? And in this way we can definitely have a good idea of what can be the target with series of KPI like kind of IT resources, resource usage, carbon footprint, kilowatt-hour consume, etc. Etc.

Speaker 1: 9:17

And what is, in your experience, the most common scenario or quick wins to be more sustainable? What do you see the most that can be optimized?

Speaker 2: 9:26

Okay. So again, if we talk about IT layer or data center facility layer, this will not be the same In terms of IT layer, of course, this is to go to, let's say, the most modern IT kind of infrastructure. So meaning the IT refresh is definitely a must. Then IT consolidation, just physical consolidation, virtualization is also a must, of course. Then, if you move to the data center facility, we got a lot of projects of data center retrofit, for example. So retrofit is always in order to implement more modern kind of facilities in terms of cooling infrastructure, and usually I mean more than 90% of the problem. This is not power feed, you know, of course you get some losses on the power distribution, but I mean the efficiency ratio of UPSs, of all those switch gears and things like that. This is not a big deal. This is not at stake. What is at stake? Definitely the cooling. So the cooling technology is really important.

Speaker 2: 10:33

And now, with the modern use case, all the customers that are transforming on the IT, they also go to more modern use case, for example, the adoption of artificial intelligence in everything. So artificial intelligence, usually when you talk, let's say, to the public globally, this is known with LLM, you know, generative AI and so on, because this is what the public knows. You know, like you know, chat, gpt just talk about this one, because this is most famous, this kind of thing. But then now artificial intelligence is something that is more and more adopted to improve the business application and the customer experience. So you get AI on everything.

Speaker 2: 11:18

I don't know if you recently entered into a cell phone shop. You go to a cell phone shop and you ask for whatever cell phone from Apple, from Samsung. So I don't want to be in favor of such or such brand, but usually if you talk about at least two brands, this is OK. And it happened to me for my son entering into a store from Orange in France and the guy said, hey, look at this new Samsung, this has an AI layer.

Speaker 2: 11:48

So I started to chat with the guy because, of course, ai I know a little bit, so he was not really aware of what was AI on his cell phone. But anyway, in any use case now there are layers of AI. This is in cybersecurity, this is in defense, this is now in manufacturing.

Speaker 2: 12:06

In manufacturing, we got a bunch of projects now with automotive industries, you know, to implement New IT layers on the shop floor. You know, industry 4.0 to even reach a better, let's say, of course, operational efficiency, I would say of the of the manufacturing, but also to improve preventive maintenance and globally, to optimize all the key parameters of a factory, globally, you know, and for that there is an AI engine on that. So this is all area of the industry. You get that also now in 5G network deployments, you know. You get that for R&D, of course, for all kind of R&D. So I mean AI is definitely at stake for all kind of use cases. And now, if you implement AI, this means that you need definitely to run new kind of IT infrastructure, you know, featuring GPUs, featuring, let's say, last class, last generation of CPUs like Gen 12 and so on.

Speaker 2: 13:26

So this means that you will need to have an IT density in your racks that is much higher, and for that you will need new kind of infrastructure featuring, for example, liquid cooling yes, and liquid cooling will, of course, contribute to sustainability objective because by definition and this is a dna of liquid cooling, so not only to cool very high density it infrastructure, you know, this is also contributing to sustainability because this lower by far, at least by four times, the power you need to cool, because you need power to cool.

Speaker 1: 14:05

Yeah, I understand. So HPE is all about liquid cooling then for the future. But I wonder then if liquid cooling is so much more efficient for those density If you take a more traditional server and you put them fully dense in a rack, is liquid cooling also more efficient?

Speaker 2: 14:25

Yeah, absolutely.

Speaker 1: 14:28

But why didn't the industry already adopt liquid cooling then?

Speaker 2: 14:32

Okay, because currently let's say 99% okay, but let's maybe it's not 99, maybe it's 95 97% of the existing data center. They are not able, because they have not been designed for that. They are not able to implement liquid cooling and to run liquid cooling IT infrastructure.

Speaker 1: 14:53

But now they are rebuilding or modernizing the data center to adopt liquid cooling Exactly.

Speaker 2: 14:58

So now I can guarantee you all the new projects on which we are working as an engineering company or as a modular data center solution provider. We got liquid cooling in it. All the new data center will have to be DLC ready, direct liquid cooling ready.

Speaker 1: 15:17

Okay, you know, in order to implement the IT, but do you think, like in 10 years, that every server will be liquid cooled then, even if it doesn't have a GPU?

Speaker 2: 15:25

Yeah. So this will not be a big bong, you know, but this is the trend. I mean in two years, three years from now, if you get a data center that is not at least DLC ready for I don't say for 100% of its floor space, you know, but at least to be able to run some DLC equipment mm-hmm this is a nonsense, you know.

Speaker 2: 15:48

This is a way. This is a way. There is no doubt why? Because even if you take, I would say, what we would call legacy server, not AI oriented servers, you know, including GPUs and so on. Look at the Gen 12, the new generation of CPUs from Intel, from AMD, you get such a high density on the processor itself that you will get what we call pizza box servers, our ProLiant line of servers pizza box server with 1.2 kilowatt per server.

Speaker 2: 16:18

So if you want to use your rack capacity, you will have densities like 50 kilowatt per rack.

Speaker 2: 16:29

And the limit is like 25, I learned on air cooling. Yeah, we can reach, let's say, 30 kilowatt, 35 kilowatt, let's say, with a very good design, but 50 kilowatts for sure not. So definitely. We need to implement DLC to take the benefit of this kind of new processes. There is no doubt on that, you know, and we have done our own modeling and simulation in real condition in our own data centers, in our labs.

Speaker 2: 17:03

So we got, let's say, a famous slide that I'm now showing in all my presentation. The modeling is for a data center of 10,000 servers and we have compared with an equivalent, let's say processing power capacity, equivalent processing power capacity, air-cooled server from our ProLiant DL family and the equivalent in DLC. Our results show that you will consume 80% less power to cool dollars, to cool in terms of tons of CO2, because this is really proportional. Also equivalent at some percentage comparison in terms of rack space. Also, and this is definitely important, because less rack space, something like from 1 to 4 in terms of rack space, means less square meters, less investment in terms of data center space and in more of the countries like Western Europe, but also in Asia, you know perfectly in Amsterdam, in Paris, in.

Speaker 2: 18:13

London in the main cities like that, not to talk about Korea or Singapore. The price of the square meter, you know, is definitely huge and is increasing. So if you can guarantee the same let's say, computing power on a space that is reduced by four, this is an incredible profit, an incredible advantage.

Speaker 1: 18:38

Okay, but is the biggest benefit to gain on the data center facility side and on the cooling side, or does the hardware configuration also play a huge role? Because we all know that AMD makes CPUs that are more power efficient than Intel because of their chip design choices in the past. But it also means that the AMD chips get less hot than the Intel ones. They still get hot, but not as hot as the Intel ones. So when you need to give a customer advice, do you prefer one chip over the other because of their design?

Speaker 2: 19:17

Usually what we will take into consideration and this is the role of our IT pre-sales engineers specializing in that. This is also in this kind of advisory, consulting, as I said in the early stages of the project, to analyze really what the customer wants to do, what is his business case, what is the various use case he will want to run, because in some cases, just let's say on a technical standpoint, if we take this key performance indicator, so the Intel or the AMD chip will be better. And also, when we talk about the chip, depending on the configuration of the chip, for the same kind of chip you can have 10, 12 different configurations in terms of memories that it uses, etc.

Speaker 2: 20:04

So, this means that definitely, the choice of the processor will be the direct result of the analysis of the application use case of the customer. Okay, so I mean we work equally with Intel and AMD, you know? So in this way we will not say to the customer okay, please select Intel, it's far better than AMD.

Speaker 1: 20:28

No, this will depend on its use case okay, and I talked to AWS a lot in the past and they introduced their own graviton ARM processor and they always said that the power consumption is so much less than a comparable model from Intel and even AMD that the savings and the more sustainability we can reach with those chips is so high. I know there aren't many server chips based on RAM, but is it something like HP is interested in, because I can imagine, especially for sustainability goals?

Speaker 2: 21:03

an.

Speaker 1: 21:03

ARM is getting more important in the data center world.

Speaker 2: 21:06

in the end, yeah, but we have a couple of HP labs program currently under development labs program currently under development, so we cannot say much about that. But typically this is about in-memory computing and this is about photonic computing. So this is not short-term programs. You know, this is program that will go.

Speaker 1: 21:30

Let's say, in the next five to ten years. But you could also uh use the umpera ultra chips. You can buy them off the shelf as well, but you're not doing that right Currently.

Speaker 2: 21:38

We are not doing that for the moment.

Speaker 1: 21:40

No, no, no but are you interested in that development in the REM space as a sustainability director?

Speaker 2: 21:50

Personally, I'm interested in everything that will contribute to sustainability.

Speaker 1: 21:54

The thing is that you know.

Speaker 2: 21:55

I'm not responsible for the processor strategy from HPE, so this is not me to decide about the adoption of such or such processors. So I know we are looking at that, but for the moment there is no.

Speaker 1: 22:10

It has a lot more technical implications if you would offer that compared to the x86 I can imagine, but. But from a sustainability point of view, you can make maybe a bigger step, and I just wondered how much impact something like that should have on the total sustainability plan you create for your customers no, definitely.

Speaker 2: 22:34

This is something that we are looking at. You, I'm interested in that. I'm reading newspapers I've been in specific technical newspapers about that but for the moment, this is not adopted by HPE for our IT infrastructure.

Speaker 1: 22:46

Okay, and if we look at the go-to-market strategy from your point of view, you have a pre-sales coming up with a configuration for one of your customers, but they want more sustainability and then you and your team try to help them optimize the configuration and the data center.

Speaker 2: 23:06

So we got, let's say, internal processes that make that each time we go for, let's say, of course, an enterprise customer In our services organization, we do not address the SMB, you know, small and medium business. So for enterprise customers, you know, we are systematically now embedded with the IT pre-sales organization in order to determine and to look at what will be the impact in terms of facilities. This is always mandatory for us because this helps a lot the customer, but also the community.

Speaker 1: 23:45

But will you change, like a configuration Pre-Sales came up with, to add liquid cooling or to use a bigger enclosure or something to get better airflow?

Speaker 2: 23:55

Absolutely, and I mean we are advising, we are the advisor for the IT configuration. If the IT configuration is something that for the IT pre-sale would be the optimal, we can come with a different conclusion in terms of the impact on the facilities and the impact on the sustainability. And the same for the labs. I mean, I have people from my team, you know, working very closely with the labs. When those you know very smart guys you know are inventing the server of tomorrow, you know we are working with them to say, hey, guy, be careful, this kind of thing, you know, in a data center this is almost infeasible or this will be so many technical challenges that we should look at another way of doing it.

Speaker 1: 24:40

Okay, I think you're in a world that's developing really fast and I think your job changes every year because yeah, and this is also definitely one of the major issues that we face with our customers.

Speaker 2: 24:53

When we were designing data center 20 years ago, we were designing a data center to be run for 20 years, 25 years, so right now, and this is why modular data centers typically are more and more at stake and interested from know, interest from the customer, because I mean, they consider now that the facility and the modular data center will have the same lifecycle than their IT equipment.

Speaker 2: 25:20

Okay, so, and for three to five years, you know typically and I mean this is not a secret, we can. We can talk about it because it was announced by the customer in the newspaper we have won very recently a very large AI program project for the French Ministry of Defense that is called AMIAD. We were in competition with Atos Bell and we won. At the end of the day. The customer clearly stated and this was, by the way, mr Lecornu, the French Minister Defense, who said in the newspaper, on the TV we know that this system will be, let's say, over in the next two years. So the life cycle, the lifetime of this very big AI cluster, this is two years.

Speaker 2: 26:11

So what did they decide for the data center? This is to implement our pods. So the IT infrastructure will be in containers in a modular data center, because in two years we will take a crane and we will remove the IT cluster and the pod from the site. Let's say, replace it with the new one. You know that will be top in class, best in class in two years from now. And this is really, you know, some change. You know this is a paradigm change from the customer perception of what is a data center, and this is why modular data centers, this is in the world the kind of technology for data center that is the fastest growing, plus 25 percent per year yeah, yeah, and the danger is because the trend was to try to keep hardware alive longer.

Speaker 1: 27:06

I had to use it for five, six years and try to extend the lifetime. With AI, there is a big risk that it's going to be replaced faster because the development of hardware is so fast and the energy gains and the cost gains will be more prevalent in that.

Speaker 2: 27:23

No, no, absolutely. And the risk in terms of sustainability if you get something that is really sustainable in operations because you've got very low PUE, green energy, very good carbon footprint in terms of lowering the carbon footprint, but at the end of the day, if you replace the whole stuff every two years. This is another issue in terms of carbon footprint, Because you need to think about recycling.

Speaker 1: 27:48

It's not so sustainable. Huh Well, it can be and typically modular data center.

Speaker 2: 27:53

You know, we have now, you know, a line of factory to when we, let's say, get back a pod, what we call a pod and this is our modular data center brand from HPE, the HPE pod. So when we take back a pod from a customer, we send it back to our manufacturing operation in Netherlands, by the way you know, at Contour in Varsevelde, and we retrofit it, we refurbish it. So now it's very well adopted. So also facilitate something that you cannot do with a brick and mortar data center. It's not easy, you know.

Speaker 1: 28:28

Okay, Pascal, I think we can talk for many more hours about this topic, but we have to wrap it up. I want to thank you for joining us at TechSign Talks on Tour and we'll be back next week.

Speaker 2: 28:40

Thank you, thank you.