Lenovo’s Scott Tease on the Impact of Energy Regulations on Data Center Design and Management

Hey guys. Thanks Withrow. We're here with Scott Ts, who's general manager for the Infrastructure Solutions Group at Lenovo, and we're talking about how new regulations are being applied to the way energy is consumed in data centers, most notably in California and probably New York, and heck and a lot of other places around the world.

Scott, welcome to show. Thank you very much, Mike. It's really good to be here with you.

Thanks for having me. Data center folks like things to be consistent and when they're not, they get a little GY and well, there's change in the air and there's a lot of requirements now around energy and it's not always clear to me that that was always at the top of everybody's mind, no matter how much we talked about sustainability in these recent years. But now increasingly it's gonna be a compliance mandate.

How do the folks who run data centers kinda wrap their heads around that and get prepared for that? Because there's slightly different nuances, I imagine, in all these different regulations. Yeah, you know, we're, we're talking about New York and California 'cause we're, you know, we're here based in the US but this is not new.

We've been seeing these kinds of rules come out. The European unions has been working on it for quite some time. Germany probably one of the leading, you know, players around the world that's been focused on how you drive a more energy efficient infrastructure for, for around it.

Uh, so we're, you know, we're catching up here a little bit, so I I kind of welcome like some of this new rulings outta New York and California because look, um, all this power that these data center use, um, to generate all that power, it, it, it's burning fossil fuel. That means we're, we've got a carbon impact on the environment. Uh, so it, it does have, you know, a direct impact on the planet.

Um, in addition to that though, it's also a big part of the, a company's bottom line. Um, power is getting more and more costly every single year. And if you can save on power and save on your energy bill while simultaneously making a big impact in a, in a good way on the environment, it's goodness.

And I think that's good behavior that we should try to, you know, we should try to push as much as we can. It seems like there's three aspects to this. One is a lot of that folks are running older servers that probably are not nearly as energy efficient as they could be and certainly don't perform as well as they probably could.

So, um, how many folks are kinda kind of be upgrading server and storage systems because of these regulations and for that matter, are they running a little behind schedule? Yeah, you know, I, I think Mike, what's, what's driving a lot of the focus, let's start with the focus, if you don't mind, on why we're seeing these rules. It's like, you know, all the new IT that's going into place, that's running high performance computing and artificial intelligence is doing amazing things.

I mean, when I think of some of the goodness that we get out of our HPC systems like hurricane prediction, things like that, knowing where hurricane's gonna land five days in advance, you know, that was unheard of a decade ago. We can do that now thanks to it. But it is power intensive.

It's very power consuming and, you know, AI is, is driving a great deal of power. As we're building out these initial kind of AI systems, we're sort of brute forcing all this advancement we're making in ai and it's, it's meaning a lot of power consumption and it's put a lot of local towns and municipalities that have limited amounts of power resources. It's kind of put 'em back a little bit and they're trying to figure out how they react to it.

So it's not only the companies themselves, but it's the local, it's the local areas, you know, data centers. Everyone would like to have a data center, you know, in their area, but it does consume a great deal of power, making sure we're managing the powers efficiently as we possibly can. It's, it's a really big deal.

And if you think about most data centers today, most of 'em are air cooled, about half the power that goes to that data center is, is is likely not going to running the it itself. It's going to running things like air conditioning and air handlers and these inefficient devices that we can make big improvements on. So again, we, we sort of welcome the, the increased focus and I view it as a necessity for the future 'cause we're just not generating enough power and we're definitely not generating enough green power for the demands that we see coming down the coming down the road.

And to your point, it would seem like there's opportunities to improve the distribution of electricity within the data center as well to make that more efficient. Is that something we need to be thinking about? Yeah, I think, you know, we're gonna have to rethink a little bit of everything of how the data center, how the data center functions.

You know, today the, you know, we, we bring in traditional high voltage power, we convert it down to something more like we're used to in our homes and try to distribute it that way. There's more efficient ways to do it distributing at, at higher voltages. That's a, you know, a way to take out a few percent of, you know, losses from energy efficiency losses we see there.

I think the bigger, the bigger consumer of power in the data center is just movement of air. Our server environment, our storage device, they all, they all like run on moving of air. I've gotta get heat away from parts and then do that.

I've gotta put fans in my servers. Those fans consume power, then I move the heat out of the server into the data center room and I've gotta get it away from the racks before they overheat the racks. That takes air movement.

Um, the thing that I think is hidden from most people is how power intensive moving air is. It's just incredibly power intensive just to move a bulk of air and imagine a data center is moving air constantly 24 hours a day. It adds up to a lot, a lot of power consumption just moving that hot air around, chilling it off and then pumping it back to those fronts of those racks again.

Uh, you look at things like our, our liquid cooling, our Neptune Leno, Neptune Liquid cooling. It's, it's, it's primary focus is to replace all that air movement and instead of moving massive amounts of air around the building, let's move a small amount of liquid liquid's much, much better at heat transference. So let's use that medium to get rid of the heat rather than air.

And I think that's our, that's our biggest area that we can really make a big improvement in how data centers operate is rethinking that whole cooling infrastructure. So I'm old enough to remember when data centers were water cooled and there was pipes everywhere. So is this kind of back to the future?

Yeah, it's uh, kind of funny. I, you know, we like to, to think we're reinventing the wheel all the time, but maybe it's just a better looking wheel. Uh, you know, the mainframes have been doing water cooling for a long, long time.

To be honest, when we started doing liquid cooling back in 2012, believe it or not, we did our very first warm water cooled supercomputer, um, oh, you know, 13 years ago now, believe it or not. Um, the engineers that did that for us all came from the mainframe. Uh, we were all at IBM at the time and it was IBM mainframe engineers.

We borrowed some of 'em and they helped us design out that first warm water cooling system because most of the rest were engineers that we were working with at the time. They had come out of PCs and the concept of water and electricity in a PC environment was kind of scary. But for the mainframe folks it was like, you know, the normal way of doing business.

So it, it's kind of a return to, I guess a return to the, to the smart way we started with mainframe. Are there also things that we can be thinking about to improve the grid itself? 'cause all these data centers are hooked up to some grid somewhere and the grids are hardly the same everywhere.

There's a lot of uneven distribution here, but it seems like we haven't really upgraded the grid infrastructure in quite some time. Yeah, that's a man, that's a really, that's a really fun topic to explore and that one will take us to full time just to talk through that one. But, you know, you think of the complexity of the grid is not only about getting power to where you need it, but it's also the mix of power that customers wanna see today.

So, you know, getting all your our, your power from coal today, for most, most companies, not an acceptable way to get it. They wanna see a mixture of different types of power. I want, I want renewables as a percentage of that.

I'm fine with some nuclear, I'm fine with a little bit of fossil fuel type fuel, um, power power generation. But I want to be able to have predictability in what I'm able to get and what I'm able to access. And then as I start looking towards regulations, a big part of how we're gonna meet those carbon initiatives in that is knowing that we're using some renewable power sources.

Um, that that's gonna be a big way that we're gonna re reduce carbon. Even if our power consumption doesn't go down, we're at least generating that power with a green source like hydro or, or solar. It, it at least minimizes the impact environmentally.

And that that's, I think that's one of the big aims that these, uh, local municipalities are trying to drive, is just making sure as we put these data centers in, they're good for the community, they're good for the environment and they're good stewards of the limited amount of power that these, uh, these local areas can actually produce and get ahold of. Do we also need to look at the software that we're running on these machines? 'cause you know, at least in my experience, a lot of it isn't very efficient and it certainly wasn't written with an eye towards reducing the amount of energy that might be required to run it, but I feel like it's overlooked area.

Oh, without a doubt, man. We are, we're in a pretty magical time, especially when it comes to ai. I mean, we're, we're at the beginning of something really amazing that we're gonna be able to do with this new kind of it, this new kind of technology.

But there's no doubt whatsoever that we are in the brute force, uh, portion of like the development curve. We're just throwing hardware at these problems and you know, it's all in an effort to go build something that we've never been able to build before, but it's not as efficient as as it's gonna be or as it can be. Um, and you know, we're, we're doing all we can to kinda lower that barrier of entry so that you can run real powerful AI on any kind of device.

You know, one of the things we're, we're, we're big on here at Lenovo is, uh, you know, technology for all, uh, bringing AI to all. Um, and a big part of that is making sure that not everything that runs on AI has gotta be on some super power intensive, you know, uber expensive machine that it can run on things like my Motorola phone right here, my think pad. Um, it can run in very lightweight devices and I can still get super powerful AI out of those devices even though they're not, you know, massively power consumptive.

A big part of that is gonna be the software, the software environment, the ecosystem, the, all the goodness that we get. You know, as you mature, you know, you mature like an it an IT space. High performance computing 15, 20 years ago was the same exact way we were throwing hardware at problems not being very, uh, smart about it, not being very elegant in the way we solutioned it out.

Now that's come a long way. We get a lot of work out of very small amount of power and high performance computing environments, but we're gonna see the same thing from ai, but it's gonna take a little while. It could take as long as a decade actually to really feel good that we've, we've started to be as efficiently as we we could possibly be on the software side.

And also we seem to be a little obsessed with GPUs these days, but I think there's other classes of processors that might be able to run some of those AI workloads a little more energy efficiently, shall we say, or at the very least, maybe even less costly. But, um, do we all understand that or are we still kind of thinking, you know, GPUs was the answer. What was the question?

Yeah, that's, I like the way you put that. It's definitely, yeah, there's no doubt GPUs have taken center stage on it because they, they are the devices that have unlocked this potential and they're gonna continue to be part of the really big AI that we do. The foundational model creation, you know, that that tier one kind of stuff that the really big players like, like Microsoft, like AWS like, you know, Facebook are doing, that's gonna be at the forefront of what they do as they build out those core models.

Now, as we get those models, especially when they're open source pri privatizing those, adding our personal data to them is a much, much lighter weight functionality. In fact, a lot of that can run not only on like a Xeon or a Epic processor, but gonna run on stuff that's much lighter weight than that, an arm ship, something like that. So that's a big focus that we've got here is right sizing the IT itself.

So that one, it can run the AI the customer needs, but it can also fit where they need it. You know, a lot of the data that we're processing is not in a cloud somewhere. It's not out in a data center somewhere.

It's here in this building like around me. I wanna be able to process that data right here, not have to ship it off some distant data center, some distant cloud, but actually process it right here where it's being created. Um, to do that, I need that.

I need that device to fit in the environmentals that I've got, the power envelope that I've got and I need it to be affordable. And that's a big part of the focus we've got here at Lenovo is just making that the, the reality. So you get powerful ai, but it's, it's, it's achievable and attainable for everybody.

So what is your best advice to folks as we look at all these regulations? What should they be thinking about? How do I kind of future proof my environment?

Yeah, I think the first thing is get a good baseline. Uh, that's the biggest guidance I give people all the time, our customers is baseline where you're at today, where's your power coming from? How much are you using?

Um, start that as a baseline. There's no magic magic trick. There's no magic bullet that's gonna solve this problem.

It's gonna be incremental improvement over and over again. Uh, but it, but together they can make a pretty significant impact on the amount of power you consume. Uh, so good baseline's, a start managing, you know, kind of step by step, how you take it lower and what steps you take is number two.

I would say rethink how that data center works. Don't be afraid to rethink that old air, cool data center. Um, a lot of times we can retrofit ancient data centers with water cooling and make 'em last longer than customers ever thought possible.

Instead of like building a brand new data center, let's talk about what we can do with what you've got by converting it from an air cold center to a water cold center. And people are really surprised how simple it is, how inexpensive it is, and what, how long we can make an older data center infrastructure go. Uh, so I think that's a big part.

And the last thing I'd say, Mike, is keep in mind that when you do something that's good environmentally, most likely, you're going to also see a business positive impact. Again, power consumption has CO2 impacts reduce it. Your CO2 goes down, but your power bill also goes down.

And power, power costs per kilowatt are not coming down. They're only going, they're only going higher. So whatever steps you take these days in the market, whatever you're paying for your power, it's only gonna get better as we, as we go into the future.

Alright. I think the equation that they came up with back in the day was, uh, energy equals mc squared. Well, I think that m stands for money that thanks.

I like it. I like it. Scott, thanks for being on the show, Mike.

Thank you so much. Take care. All right, I'm back to you guys in the studio.