A Better Life: The Case for Cloud

Jolie Hales:

It’s like suburban, rural, I would say. It’s not quite rural.

Ernest de Leon:

Right.

Jolie Hales:

Did I say suburban? Yeah. Suburb. So why did I think about a car; Suburban? I wonder if that’s where they got that name.

Ernest de Leon:

I think so. You needed that big of a tank to get out there.

Jolie Hales:

Hi everyone, I’m Jolie Hales.

Ernest de Leon:

And I’m Ernest Leon.

Jolie Hales:

And welcome to the Big Compute Podcast. Here, we celebrate innovation in a world of virtually unlimited compute, and we do it one important story at a time. We talk about the stories behind scientists and engineers who are embracing the power of high performance computing to better the lives of all of us.

Ernest de Leon:

From the products we use every day to the technology of tomorrow, computational engineering plays a direct role in making it all happen, whether people know it or not.

Jolie Hales:

Ernest, oh my gosh. I feel like I haven’t talked to you forever. I don’t know. Remember that time we used to publish new episodes every couple weeks, and then we didn’t for like, forever?

Ernest de Leon:

Yeah, I believe the pandemic was still a thing back then.

Jolie Hales:

Yeah, probably was. I think it’s safe to say it’s been a little while, but in case our listeners are wondering, if you’re still hanging around, we are still here. We’ve just been super swamped. We had a big successful Big Compute virtual conference in November.

“Here, you are part of a cloud-first high performance computing community of science and engineering thought leaders who are building the future with virtually unlimited compute.”

That had Joris at Rescale, and Jensen Huang from NVIDIA, Sir Richard Branson of Virgin Orbit, before Virgin Orbit kinda… had their final orbit.

Ernest de Leon:

I believe in space parlance, we call that rapid unscheduled disassembly.

Jolie Hales:

But it was a great conference and anybody interested in seeing any of the videos and the presentations, you can go to bigcompute.org. I totally recommend the one from NASA, by the way. I don’t know if you’ve seen that one, Ernest, but it was awesome.

Eric Nielsen, NASA:

Can we bring exascale class computing to bear so that we can do higher fidelity modeling and simulation to, perhaps, take that off of the flight plan, so to speak, and potentially save a company several hundred million dollars in flight testing by doing it virtually inside a supercomputer.

Jolie Hales:

There’s also a highlights version of the keynote if you have a short attention span, as many of us do these days, and that has Joris and Jensen in it.

Jensen Huang, CEO of NVIDIA:

With NVIDIA in every cloud, researchers and engineers can instantly access elastic GPU-accelerated computing from their browser. Rescale’s cloud-based platform enables customers to run simulations on demand and scale up or down as needed.

Jolie Hales:

Beyond that, I think it’s safe to say that a lot has happened in our personal lives. When we last talked, Ernest, I think your wife had just had baby number two and a surprise early delivery and now he’s what, like a year old? Oh my gosh, has it been that long?

Ernest de Leon:

Yep. He’s actually 13 months and a complete terror on all fours crawling all over.

Jolie Hales:

Yeah. How’s that going?

Ernest de Leon:

It’s intense. Just absolute craziness. What about you?

Jolie Hales:

Oh, yeah. I mean it’s the same story. You have one kid, you’re super busy, but you can at least get things done during nap times. But when you have two kids, there’s just like no relief, right?

Ernest de Leon:

Yep.

Jolie Hales:

I think when we last recorded, I was in my third trimester, and I remember in the last episode whining about gestational diabetes.

“And that basically means that I can’t eat any sugar, and I can only seldom eat carbs. And I’ve got to tell you, Ernest, it is the worst.”

And for those who were wondering, I’m happy to report that baby number two, second boy of two boys, I can say that thanks to insulin and the tortures of a healthy diabetic diet, he was eight pounds, 13 ounces. And while that may sound large to some moms, that is super small for me. My first boy was 10 and a half pounds, so this was progress.

CLIP: Michael Scott, The Office:

Man, was he fat. So, so fat.

Jolie Hales:

You and I lived in the same state, too, when we last recorded. Not that we were close by to be able to have play dates with our kids. I think we’re still like eight hours apart driving distance.

Ernest de Leon:

Yeah, I was going to say, I think the flight time between the two is the same.

Jolie Hales:

Yeah. It might actually be faster, now. I don’t know. But yeah, that’s because, for our listeners, I uprooted my family after living in Southern California for 18 years, and I actually moved them back to northern Utah where my husband and I both grew up.

Utah State Song:

Utah, people working together. Utah.

Jolie Hales:

And where our kids have lots of cousins and aunts and uncles and grandparents and all the fam-bams around. And I will say that while the weather is taking some adjusting, there was way too much snow here for a couple months, and I also hate the mosquitoes.

CLIP:

Ah, mosquito.

Jolie Hales:

Because there are billions here, where Southern California doesn’t have bugs. It’s otherwise been a really awesome change. We exchanged a backyard view of a Burger King–

CLIP:

Can I take your order?

Jolie Hales:

For a view of horses and chickens.

Ernest de Leon:

Yeah, California’s definitely a different place. My wife and I spend it with our family, we spend a lot of time here, but I would love to permanently move back to Texas. As a matter of fact, I bring it up with my wife every now and then. We have a ranch south of the city where we grew up.

Jolie Hales:

That sounds amazing.

Ernest de Leon:

Yeah, and I keep thinking about it, and my wife disagrees. She’s like, “Well, I don’t want to move back to Texas, and rural life is quite different.” I’m okay with it.

Jolie Hales:

Totally different.

Ernest de Leon:

But the real deal is, it was late March, early April of this year, we went back for an alumni event at the university we both attended, and Texas starts to warm up in April, it doesn’t get to its peak heat. And we thought, let’s take our kids out on this thing San Antonio has there called the River Walk. Both of them were just red in the face and-

Jolie Hales:

Oh, no.

Ernest de Leon:

Very unhappy with the level of heat. They’re just not used to that because here, in the Bay Area, it gets hot,

Jolie Hales:

It’s beautiful.

Ernest de Leon:

Don’t get me wrong, but it’s nowhere near what Texas has, so.

Jolie Hales:

Oh, totally.

Ernest de Leon:

On that aspect, I agree with my wife that it’s all about the weather here. It really is.

Jolie Hales:

Well, I moved to Southern California, originally, for Hollywood. I got my masters in film directing there. It’s the place to go for that kind of thing, but I stuck around as Hollywood became more dispersed and you were able to work from different places, I stuck around because of the weather. Let’s be honest, the weather rocked there.

Ernest de Leon:

Right. My wife and I laugh every time. It gets to the summer around August here, and the news is all about there’s a heat wave.

Jolie Hales:

And it’s like 85.

Ernest de Leon:

Yeah, it’s going to reach 92 degrees, and we’re laughing about it. We’re like, that’s a nice Texas spring day.

Jolie Hales:

Right.

Ernest de Leon:

But yeah, the people here, especially the ones who are born here.

Jolie Hales:

I was going to say, yeah, the people–

Ernest de Leon:

They can’t handle it.

Jolie Hales:

Yeah, the people who appreciate California weather the most, locals who’ve lived there their whole lives appreciate it, but those of us who lived or were raised in other places with these weather extremes and then relocate to Southern California, I literally walked outside every day for years, probably the, it couldn’t even be the full 18 years, I just was like, oh my gosh, this place has the best weather.

Ernest de Leon:

The problem is just the traffic.

Jolie Hales:

Yeah, there’s so many people.

Ernest de Leon:

There are too many people in too small of an area.

Jolie Hales:

And that’s the nicest change about relocating back to Utah. And we went to, I would say it’s not quite rural, we have enough established neighborhoods that there are some HOAs. We actually do have an HOA, but whatever. But we do have fiber internet, which is great because my internet’s so much better here than it was in California, believe it or not.

Ernest de Leon:

Oh, I believe it.

Jolie Hales:

But anyway, I suppose we should shift away from our narcissism and toward relevant technology topics here, at the Big Compute Podcast. And to start us off, I actually have a question for you, Ernest. So how familiar are you with the term “server hugger?”

Ernest de Leon:

Not much, other than I heard they staged a protest against some company named Gone Prem a few years back.

CLIP: Clifford, Gone Prem:

Call Gone Prem today, and you’ll be on your way up to the clouds. Kaboom.

Jolie Hales:

In fact, kicked the founder of Gone Prem off of campus at SC19, if I remember right.

CLIP: Clifford, Gone Prem:

I’ll recycle all your old supercomputers that you’re not taking up to the cloud.

Jolie Hales:

I know we have mentioned this term before on the podcast, but for listeners who aren’t familiar, a server hugger is a term used by cloud computing advocates, which Ernest and I are, if you haven’t figured that out by now, and it’s used to describe someone who doesn’t want to move to cloud. So this server hugger would want to, instead, maybe keep all of their hardware on prem and in house.

And there are a number of reasons reported for this fear of the cloud. So I wanted to go through some of these. First, companies are afraid that moving to cloud will mean that their employees will have to change the way that they work. And I’m going to be honest, I think that’s a valid fear in a lot of ways. Change is hard on people. People don’t change because obviously, it’s uncomfortable, but then again, I guess my rebuttal to that would be to say that change can also, obviously, bring good in a lot of instances.

For example, as a filmmaker and a podcast producer, myself, I constantly have a lot of projects in motion and in media like that, there’s a ton of different elements and tasks involved in those projects. And I used to do all of my project management in a Google spreadsheet and I had this process set up and it worked fine for me, and then a new boss came along and actually challenged me to migrate to the project management platform called Asana, which I actually think Sam Altman is an investor of Asana.

CLIP: Sam Altman:

I’m a big believer that real sustainable and improvements in quality of life come from scientific and technological progress.

Jolie Hales:

But while it was this painful month-long process to move everything and change all my processes when I migrated to Asana, the end result was basically, I can say, life changing. I love Asana. They’re not a sponsor of this podcast, but I will sing their praises because I love organizing things in Asana. All my projects, now, are streamlined better, my processes are better, my communication is better, and that’s all because I changed the way that I worked. But I may not have made that change without having a boss telling me that I needed to make it. And, I would say that that same concept applies to a lot of different technology, but in this case, moving workloads to the cloud from on-prem.

So yeah, it’s probably going to change the way your employees work, but is that necessarily a bad thing? What if the end result is that the clutter is removed or projects are accelerated, the human working experience is better? Yes, there’s going to be some bumps in the road along the way and it might take some adjusting time, but then it could be more efficient and better in the end, so shouldn’t we all be striving for beneficial digital transformation? And I know that in the example of Rescale, who is our podcast presenting sponsor, moving to the cloud means that all the annoying parts of running workloads are basically streamlined so that an engineer can just log in and focus on innovation.

Ernest de Leon:

So if you think Asana is great, wait until you use Jira. Have you used Jira before?

Jolie Hales:

No, but I keep seeing that your team has switched over to Jira, but I don’t really know anything about Jira. I think Asana’s geared more toward marketing creatives, whereas, Jira might be geared more toward security and developers. Is that true? I don’t know.

Ernest de Leon:

I think that’s a fair way to look at it. I think Asana is more geared towards less technical users.

Jolie Hales:

Okay, well, that would be me.

Ernest de Leon:

And I don’t mean the person’s skillset is not technical. I mean what they do, their job.

Jolie Hales:

The tasks.

Ernest de Leon:

The tasks are less technical.

Jolie Hales:

That makes sense.

Ernest de Leon:

Jira is meant for, basically, engineers.

Jolie Hales:

That makes sense.

Ernest de Leon:

But I’ll agree with the overall sentiment you painted here, which is moving from an older, more cumbersome way of working, which all of us have had, at some point, with inherent limitations, because that’s how those things happen, to a newer, more streamlined way of working can be awesome. And not only that, but as somebody who’s been doing this for over 25 years, at this point, I can tell you that this industry that we work in that has to deal with high performance computing and computing in general, if you are afraid of change or change challenged, if that’s a term, this is not the right place for you.

Jolie Hales:

Right, yeah.

Ernest de Leon:

There is nothing consistent in our world other than change.

Jolie Hales:

That is such a good point. When has technology really stood still? Even with Moore’s law slowing down, obviously AI is on the rise. There’s just so much coming at us, and I think that’s a really, really good point.

Ernest de Leon:

Right, you have to be able to change.

Jolie Hales:

And it’s nice to be in a comfort zone, I think where you’re working, in a lot of ways.

Ernest de Leon:

Absolutely.

Jolie Hales:

And so it’s hard to want to veer out of that because it takes a lot more work. It takes a lot more brain power. And when you’re in a comfortable place, it’s like you’re not really looking for that, even if it’ll make your life better in the end.

So, change is hard and that’s understandably a fear for people and turns them into server huggers, but there’s another fear that’s been reported, too, that I hear a lot about when it comes to moving to cloud, and that’s the concern that maybe a company will end up paying more or maybe they don’t know what they’re paying for and they start using the cloud and they’re billed for all these things that they weren’t expecting to be billed for. And just the overall newness of cloud financial commitments, because it is a different model. You don’t own the systems.

And I guess, to that, I would just say that yeah, of course, it’s different, it’s change again, right? Because you’re not paying for that on-prem hardware anymore. It’s more along the lines of licensing and core usage and such, which is why, I guess I would say, you just want to make sure that you use a cloud HPC platform or provider that not only gives you financial controls, but gives them in a way that is easy to understand so there’s not all these hidden charges, so you’re not wasting time on trying to figure that out and you don’t get stuck with some weird hidden fees or something. And I think that’s getting better and better in the industry. I know Rescale is really simple, to throw them in there, again. But yeah, that’s one of the fears.

Ernest de Leon:

Exactly. And it all boils down, at the end of the day, to usage patterns. So selling the cloud story is based on shared computing, is what it comes down to. So if you’re already running at a hundred percent capacity, 24-7/365 on prem, and you do not ever need to exceed your current capacity, so imagine you have an HPC cluster, you’re running that thing 24-7/365 at a hundred percent utilization, okay? Then the cloud argument is tough because you already know what the costs are and you’re eeking every bit of capability out of that system, and you don’t need anything else. So for you, you’re probably fine where you are, but if you have any other usage pattern than that, so-

Jolie Hales:

Which is pretty much everybody, though, because who runs their on-prem system at a hundred percent capacity every single day and never needs to burst?

Ernest de Leon:

Exactly.

Jolie Hales:

Who is that?

Ernest de Leon:

That last part is key. There are people who run a hundred percent capacity 24-7/365, but that’s because they don’t have the capacity to burst, so they’re limited by that and so you end up in queues waiting for jobs to run. And so again, it’s exactly like you said, if you have any other usage pattern than that, which is almost everybody, then cloud makes sense because you have virtually unlimited capacity and the costing model is very clear, and the most important thing is, it is my job. So you run that thing through, you get your results, and if you do that same thing repeatedly, you have a very good idea what that costs.

And going back to your earlier example where people are afraid they might pay more because they don’t really know what they’re paying for on the cloud side, I would argue that aside from the people who have been running HPC clusters and data centers for many, many years, if you were a new business coming up and you needed to use HPC for whatever reason, and you didn’t have the history of running on-prem clusters and data centers or using the cloud, so you’re brand new to either one, both of them are a black box, in terms of what costs are, because what a lot of people don’t consider-

Jolie Hales:

That’s a good point.

Ernest de Leon:

There’s a lot of other costs that go into running on-premise systems that you wouldn’t know about. So either way, it’s a bit of a black box. And I think it goes back to just the fact that people are comfortable because they’ve been doing this for many years and they know how it works, they know what their costs are, and they don’t want to accidentally step into something that ends up skyrocketing their costs when they were fine, so to speak, with what they had before. And that’s really what it comes down to; it’s an unknown and it’s a bit scary. But like you said, companies like Rescale make this super easy for you to do.

Not only that, but you can actually cap your cost. You can say, okay, well I want to try this thing out. I want to run a couple of jobs, I want to do this and that, but I’m going to put a limit. I’m going to put a budget on this, call it research and development in the cloud HPC for my company, I’ll put a cap on that and I’ll give some engineers some access and say, “Hey, kick the tires on this thing and tell me how it works.” And I know that I’m never going to exceed that cost cap that I put on it.

Jolie Hales:

And you’ll get alerts that will tell you how much, what percentage you’ve used, so always kept in the know.

Ernest de Leon:

Exactly. And then when you roll into production eventually, because I think that’s kinda the trend there. People start kicking the tires, they use it, they realize, hey, this is way better than what I had, and so they start moving their workloads in. Understand that same cap premise that we used before for, you kicking the tires and trying it out, applies to your actual production workloads. You can put caps by all kinds of things and just make sure, so you actually have complete control of your cloud spend with something like Rescale. So that fear is a little bit mitigated. I think it’s still a valid fear in some areas because there are a lot of products out there that don’t give you the cost controls, right?

Jolie Hales:

Yes.

Ernest de Leon:

They want you to use more, right?

Jolie Hales:

Yep.

Ernest de Leon:

Whereas, Rescale was designed initially to be a B2B product, and every business wants to control costs, all of us, so that makes sense. Hey, it’s an understandable fear and here’s an understandable feature to help allay that fear.

Jolie Hales:

Well said. And then another fear, along the same lines that I’ve heard on occasion, is the fear that moving to cloud means that you just can’t go back to on-prem. You can’t go back from that comfort zone that you’ve been in. And this is one I struggle with because it goes back to the change idea, because as long as you still have your physical hardware, you can always go back to it. But how many people have gone back to, I don’t know, writing physical letters for business correspondence once email became a thing?

Ernest de Leon:

Right. How many people went back to horse and buggies once the combustion engine was a thing, right? It’s all the same thing, but you can always go back, but the cost is prohibitive, right?

Jolie Hales:

Right.

Ernest de Leon:

So one of the things with just computing in general, but specific to HPC, is physical hardware’s only part of the picture. You still need to account for a whole lot of other things, right?

Jolie Hales:

Yes.

Ernest de Leon:

Supporting infrastructure, you have physical building, maintenance and operations, you have power, you have HVAC, telco, as well as the army of people you need to keep all of that running.

Jolie Hales:

Yep, and you have to keep it updated constantly.

Ernest de Leon:

Right, so all of that costs money. And when you go with cloud, you’re essentially outsourcing all of that. You’re saying, Hey, yeah, you deal with all that. I’m just going to buy it as a service. And you make a good point if someone goes the cloud route and then for whatever reason says, you know what? I really want to go back to on-prem, then they have to think, okay, did I sell the building? Did I sell those servers? I need a rerun power, HVAC, telco. And I think the biggest one, when it comes to physical hardware, is the minute you buy something, a piece of technology, whether it’s a phone, a server, a laptop, it doesn’t matter what it is– The minute you buy it, it’s already obsolete, brand new in the box, because there are already new products in the pipeline that are much better than what you just bought that are going to come out in the next year, two years, three years.

So the nice thing about the cloud is you don’t ever have to upgrade anything. The next time this new high performance core type pops up, it just shows up in your dashboard and you use it. You don’t have to worry about ripping out a bunch of servers and replacing them with these ones with newer pro, you do none of that. It’s all done for you, all you got to do is use it.

Jolie Hales:

Right, and I think that while migration to the cloud, when it comes to high performance computing, at the beginning it was a little bit slow. Now, I think it’s pretty much safe to say that those in the industry, at this point, are acknowledging that cloud is not really only the future of HPC, but it’s the now of HPC. So it makes sense, at this point, to at least make a plan on how to move some workloads to the cloud, or at least dip a toe in the cloud pool because chances are, the industry will push many companies in that direction eventually, whether they’re ready for it or not.

So, I don’t know. I feel like if you are one that still wants to keep using on-prem, that’s fine, but I would have a cloud plan for, maybe, bursting to try it out.

Ernest de Leon:

You’re absolutely right. And there’s, perhaps, a 1% slice of the current on-prem HPC market that’s not going to ever move to a cloud-based model. And they have very specific and esoteric reasons for it. So that’s fine. Cloud’s not for everyone, but for the vast majority, they’re eventually going to move, once they try it. So when your newer scientist comes on board at your company that’s designing something awesome, and that newer scientist gets told, “Hey, you’re a newer scientist and your project is less important than these dozens of other projects we have going on that have to do with the main lines of revenue to support this company, so you can’t have time on the on-prem cluster, go use a cloud model.” That person, when they go use the cloud model, they’re going to be like, well, this is way better than–

Jolie Hales:

Why would I want to use the on-prem system?

Ernest de Leon:

And you have now just made a cloud HPC user of that person for life.

Jolie Hales:

And that’s happening more and more, in reality.

Ernest de Leon:

Right, that’s exactly what happens. And so eventually, your pool of engineers on staff that are using HPC are going to slowly be cloud native, as opposed to on-prem native. And there will come a time when it’s not just a question of dollars and cents or people’s level of comfort, it is this group of engineers refuses to use the on-prem stuff, and that’s it. When that happens, you’re done. So you move on to the cloud.

CLIP:

Your software was exported to the computational cloud.

Jolie Hales:

There’s one final fear that I hear a lot, especially from server huggers that work in IT, and it’s this fear that they will lose control because their hardware won’t be in-house and physically accessible to them anymore. And I’ve even heard some IT professionals whisper that they’re worried that their jobs will become irrelevant if they move to cloud. Probably similar to how all of us are wondering how AI is going to affect our livelihoods, as well, and I think that is a rational fear to worry about how you are personally going to feed your family, should circumstances in your industry, or even your company, change.

But the worst thing I think we could do in evolving circumstances is to deny the evolution and hope that everything stays the same, because it won’t. It never does, right? I would state that it’s best to anticipate the changes and find ways to work well along with them, at least where possible. And this is one fear, in particular, that was addressed in a recent conversation that I had that I did want to share with you and our listeners. This is a conversation I had with a man named Anand Kumar, who is-

Anand Kumar:

I’m Director of IT for Cloud Application Infrastructure and Application Operation and Maintenance.

Jolie Hales:

For the company, UD Trucks.

CLIP: UD Trucks:

At UD, we know trust has to be earned, and it’s built on actions, not words. That’s why for over 85 years, we’ve been building reliable, high quality trucks that always deliver on performance, drivability, safety, and innovation.

Jolie Hales:

Ernest, have you heard of UD Trucks?

Ernest de Leon:

I have not.

Jolie Hales:

Yeah. So I think if you don’t work with trucks, it’s easy not to notice truck brand names. I don’t know that I could name a lot of truck brands, per se, except for those that also dabble in consumer automobiles. But now that I’ve done this interview, I now notice UD trucks everywhere, and I see them instantly. So for those who aren’t familiar, UD Trucks is a Japanese owned global company that is actually part of Isuzu, and they’re known for making these sleek, flat nose box trucks, you could say, or trucks that are basically the size of average home moving trucks.

Anand Kumar:

Since its inception in 1935, our company had been innovation leader with a clear vision. We provide the truck which the world needs, and it’s always committed to go extra mile for our smart logistic solution with the most dependable solution on demanding customer.

Jolie Hales:

They’re sold here in the United States and in more than 60 countries, but their main markets are in Japan, South Africa, Australia, Asia, and the Middle East. So our listeners in those areas are likely to see them more often, although I have noticed them here in Utah, actually, even though there aren’t any UD vendors in this state, apparently. I still see the trucks. And when I asked Anand what sets UD Trucks apart from other truck designers, I thought he might say maybe efficiency or capacity or something like that, but nope.

Anand Kumar:

I can say that it’s a better life. So it’s our core value. Better life is our purpose. Better life for ourselves, better life for business, better life for universe, better life for our people, and for every region, better life is our only thing why we exist.

CLIP: UD Trucks:

Driven onward by our purpose, better life.

Anand Kumar:

We provide the truck and services that the world needs today, and we are committed to make a better life. And I said that we are seeing a better life on every individual right from the top to our dealers, to our factory, and we believe in making life better for the people, planet through sustainable transport solutions.

Jolie Hales:

A better life.

CLIP: UD Trucks:

For a better life.

Jolie Hales:

That’s the goal and the mission of UD trucks, which, how can you argue with that?

Ernest de Leon:

Absolutely. If a truck can give you a better life, sounds like a deal to me.

Jolie Hales:

Right? And I could say that my pickup truck gives me a better life because it’s pink.

Ernest de Leon:

I can’t wait.

Jolie Hales:

Did you know that?

Ernest de Leon:

You told me about it. I can’t wait to see it when I’m up in the great Salt Lake.

Jolie Hales:

That’s right, we have a leadership summit for our company, so we actually going to meet in person for the first time, and you will get the privilege of meeting my 2022 Toyota Tacoma that is wrapped in pink. It’s amazing.

Ernest de Leon:

I’m still driving an almost nine year old Jeep, so.

Jolie Hales:

Yeah, but the Jeep is great.

Ernest de Leon:

Oh, the Jeep is great. It’s just old.

Jolie Hales:

And you have a Tesla, and you’re going to get a Cybertruck, and then you can wrap that in pink.

Ernest de Leon:

Yeah, I don’t. Part of me is not believing that’s ever going to happen, but we’ll see.

Jolie Hales:

Oh, no.

CLIP:

Pink is my favorite crayon.

Jolie Hales:

So Anand says that the way that UD Trucks focuses on this mission of a better life is by really diving into what the needs of society are, and then creating products based on the needs of society, which sounds simple, but I think there are companies out there that make a product that they want to make, and then they try to convince the world that they need it.

Anand Kumar:

We’re going extra mile for our customer every single day.

Jolie Hales:

And that’s one thing I love about hosting this podcast is that we get to talk to people from so many different industries whose goal is, literally, to make the world better, and in ways you would never expect, but that are totally relevant. Anyone who has bought a new car and compared it to the one they owned in the eighties will tell you that life has gotten better on the roads, just based on that vehicle alone.

CLIP: 80s Car Commercial:

But then, once you start driving it, who wants to stop?

Jolie Hales:

And that’s the case because of all of these undercover superhero engineers and researchers, as we say, who are constantly improving products.

Ernest de Leon:

Yeah, the eighties were undoubtedly the best decade ever. But there were things, like the automobiles, in that era that were just horrible, unbelievably horrible.

Jolie Hales:

Except for the DeLorean.

Ernest de Leon:

Right, and it didn’t make it. That should tell you something.

CLIP: Back to the Future:

Quick, cover the DeLorean.

Ernest de Leon:

But this is one of those things where it really comes into play in places like the Bay Area. I don’t remember how many years ago it was, but there was a Budweiser delivery truck that got stuck in San Francisco. It essentially got high centered on the trailer portion when trying to deliver beer driving inside of a city that was never designed to have an 18 wheeler drive through it.

Jolie Hales:

Oh, man.

Ernest de Leon:

Yeah, they had to call a wrecker to get it off its high center thing. And so this is where these little trucks, these, I believe the slang term for them is bobtail trucks, make perfect sense. And then similar to Amazon and UPS trucks, it makes sense for that last mile to be in a truck that’s smaller, more agile, but is still designed to carry a good amount of capacity so that you can get things to their end destination in a manner that works within the type of city that you’re driving.

Jolie Hales:

That’s a good point. And it is fun actually, I think, to watch the evolution of Amazon trucks.

Ernest de Leon:

Oh yeah, we have the electric ones here, now.

Jolie Hales:

Yeah, yes. They’ve got the neat headlights.

Ernest de Leon:

Yeah. And then that giant wrap around LED light on the back.

Jolie Hales:

Yes, yes. That’s so cool.

Ernest de Leon:

Yeah, they’re awesome. I wouldn’t say that half of the ones here are electric, yet, but there’s enough of them that you notice them all the time. Yeah.

Jolie Hales:

Yeah. And, as I spoke to Anand, it became clear to me that he’s not only the smart guy who is responsible for helping to create this better life, these better products, but he’s also super personable.

Anand Kumar:

I love to connect to people. That’s my mojo. I feel good when I connect to people. I’m adventurist by nature, so I do cycling, motor biking, offroading, lot of activity. I go, I do the Himalayas riding. Recently I finished bike trip to Himalayas, and when I go offroading, I meet with local people, I meet different city people and in different cultures.

Jolie Hales:

Anand is based out of India, and how would it be to just be able to bike to the Himalayas over the weekend? Have you ever been there, Ernest, or to India at all?

Ernest de Leon:

I’ve never been to India. It’s one of the places I would love to visit. It’s a lot more difficult, now, with small children.

Jolie Hales:

Yes.

Ernest de Leon:

But yes.

Jolie Hales:

We’ll go in our sixties. But for Anand, as the Global Director of IT for UD Trucks, you can imagine Anand has a large slew of responsibilities.

Anand Kumar:

I have 600 plus applications to take care of, out of which 70 plus are premium applications.

Jolie Hales:

So on a daily basis, Anand oversees these 600 plus applications, and then he makes sure that the IT systems are running smoothly, he focuses on vision, partners, new enhancements and developments all the way down to budgets and funding and future planning, the whole thing that has to do with IT.

Ernest de Leon:

And how many of those applications are HPC-based?

Jolie Hales:

He said about 30 or 40 of them are used in high performance computing, which just goes to show that a lot is going on in his work life, not just HPC, but IT from every angle of UD truck development. But when it comes to HPC applications, Anand’s team use them whenever computational simulation can help with design and testing.

Anand Kumar:

We do a lot of testing about our truck performance, our truck noise conditions, how they will behave in the real conditions, grass conditions, weather conditions, and we create a [inaudible] statement out of it. We do it virtually so that we can learn from the [inaudible] statement, and we understand that how they will perform with a given load in the situations at the actual, and then we take that [inaudible], break into the smaller parts, and then ask our HPC to solve it, and then give us the outcome with our server node so that we can learn from those simulations that how we can perform and size our truck or design our truck efficiently, this simulation data.

Jolie Hales:

So they’re running computational simulations for all the components that we typically see with automotive design, right? Noise conditions, weather conditions, and so forth. And one thing that they are able to do quite efficiently with computational simulation as opposed to maybe physical testing, is computational crash testing. While they still run physical tests, being able to do simulation first allows them to get a good sense of how a truck would perform under different crash conditions if it were carrying, maybe, different loads, different weights, and then they can tweak and iterate on the designs before producing the physical prototype rather than running every test as a physical prototype test, something that gets, as you would imagine, really expensive, really fast.

Anand Kumar:

It would be quite difficult if we don’t use simulations in supercomputation in real time because it will have a high cost. If you wanted to know the chassis load, we have to load the truck in actuality, if we want to know the braking and chassis braking test, we have to actually break the chassis, and it’ll cost you more than virtual computation costs. Hence, we use supercomputations to help us simulate exact result like we are doing at actual.

Jolie Hales:

It’s not just helpful in terms of time and cost.

Anand Kumar:

In actual testing, also, when you crash the truck, you have to put lot of machineries to collect the data, but with simulations, the data collection is at the click of the button.

Jolie Hales:

So without simulation, even if you had all the money and all the time in the world to crash test as many trucks as you wanted, so you crash a truck, and then what? How do you get the data from that crash test? I would imagine you would probably have to put a ton of those little sensor nodes or something all over the truck before you crashed it in order to collect real data, like, maybe, you would on a motion capture actor for a video game or something. I don’t know.

Ernest de Leon:

Yeah, you can imagine the actor who played Gollum in Lord of the Rings, and then you just smash him into the wall a hundred times.

Jolie Hales:

I don’t think he’d be a fan.

CLIP: Gollum:

You don’t have any friends. Nobody likes you.

Jolie Hales:

But if you’re crashing a person or a real truck, it’s obviously different than if you’re crashing a virtual truck in a virtual world. Since in a virtual world, everything is done computationally, so then all of the data points are instantly available digitally, leading to faster acquisition of knowledge that helps companies make cars and trucks and airplanes, and pretty much all products safer and better.

Rachel Pruitt:

This is why I love high performance computing, because it’s just so cool. Our customers are literally changing the world, and it’s so much fun to listen to what they’re doing.

Jolie Hales:

That mystery voice is another person who is part of this story. Rachel Pruitt, who is a-

Rachel Pruitt:

Product Marketing Manager for Microsoft Azure, for Azure High Performance Computing and AI Infrastructure.

Jolie Hales:

So she manages messaging and branding for Azure based out of the Seattle area, and when she’s not working-

Rachel Pruitt:

Spend a lot of time with my husband, Michael, and my new puppy, Stubs, who is five months old and is adorable and a little troublemaker.

Jolie Hales:

I’ve seen pictures of this puppy and oh my gosh, talk about cuteness overload. He is this Dachshund mix. So he has one of those long dog bodies and little stubby legs, kind of like a Corgi does.

Ernest de Leon:

Yeah, we used to call them weenie dogs when we were little.

Rachel Pruitt:

He has figured out how to jump out of his playpen, so it’s very difficult to contain the puppy because he just jumps right back out and feels like his life should be lived on the couch.

Jolie Hales:

And while we’re giving some pet love here, Anand also has a turtle.

Anand Kumar:

I call it Trudeau, but he understands my emotions, I understand his emotions, and we used to have endless conversation by staring each other.

Jolie Hales:

Do you have any pets, Ernest?

Ernest de Leon:

Unfortunately, no. I had a dog for 15 years and she left us early in the pandemic, and we are looking to get another one soon because our daughter loves animals, but she’s allergic to pet dander, I guess, so we have find a very specific breed of dog.

Jolie Hales:

Yeah, I have three. I have a 16-year-old Italian Greyhound named Comet who is blind and deaf and on his last leg, but hanging in there. And then I also have a 7-year-old whippet named JPEG, and a 7-month-old puppy whippet named Pixel. So hooray for animals, but I digress. What were we talking about? Oh, yeah, simulation instead of physical testing.

Rachel Pruitt:

We used to physically crash vehicles to see what would happen. We used to have to make thousands of prototypes for every part, every piece of every vehicle, just to get the right one because they would take it, they would test it, something would be wrong, they’d have to create a whole new prototype, test it, see what went wrong, and these are things that they’re literally just doing in the computer, now. I always compare it to almost an animated movie where you can put it into the computer and you can see how everything’s going to react.

But what makes it even cooler is that you can get data that you never could before, and some of these parts are really small or somehow, these parts work together, is really, really intricate and you can never see it in person. And so I think to these drug manufacturers that are dealing with these little tiny atoms, and I’m not a scientist in any way, shape, or form, but things that you can’t even see and are able to do these on the computer now, so well, that’s not exactly the same as here. There’s just a lot of things that you just simply can’t see in person to be able to capture that data and to make sure that vehicles are safer.

Back in the day, you got a car accident, you worried people would die. Not that people don’t die now, but it just happens so much less because vehicles are just so much safer than they ever were before. So it’s really exciting, the technology and how it advances and how it’s changed over time.

Jolie Hales:

And as we’ve talked about here, on the show, a number of times, a lot of simulations involve crunching such incredible amounts of data that it’s not possible with even high-end laptops or desktop computers that are really ramped up. You really do need the compute power of dozens or hundreds, in some cases, even thousands of computers working together in tandem to reach a solution without taking years, sometimes literal years, which is exactly why we have HPC, or high performance computing.

Ernest de Leon:

Absolutely. It reminds me a lot about making modern movies in 4K where a lot of people don’t understand the amount of compute power-

Jolie Hales:

Oh, my gosh.

Ernest de Leon:

Required to render these CG scenes.

Jolie Hales:

Yes.

Ernest de Leon:

So back in the day of 1080p movies, you fundamentally had the same applications and horsepower, and you had so much time budget to make that happen for a movie. But now that they’re shooting these things in 4k, 6k, 8k, when they go to actually output the final movie, unfortunately, the CG renders are still happening at 2k and under.

Jolie Hales:

Yep.

Ernest de Leon:

So when you look at the movie, it’s very obvious when you went from filming to CG.

Jolie Hales:

Yeah.

Ernest de Leon:

Because they don’t have the horsepower to generate that CG at a full 4K to match the rest of the movie.

Jolie Hales:

Which is so fascinating, and you’re totally speaking my language. I know that just for me, there have been a number of times where I’ve done some kind of complex visual effect in After Effects, and then I try to export or render it, and the timeline comes up and it’s like two days. And I’m like, dang it, because I’ve got a really solid machine, it’s got tons of RAM, way more than your average person, and it still will just give me this timeline that’s like, are you serious right now? And it makes me want to tap into Rescale or high performance computing. The day that I can use cloud HPC for my own personal video renders, that will be a good day.

Ernest de Leon:

And it’s all business based. It’s not that people like Weta Studios can’t double the amount of hardware to render something out in a day at 4k, it’s that the movies have to get out at a certain time. So you only have so much time and you have so many scenes you have to render.

Jolie Hales:

Yes, right, it’s the deadlines.

Ernest de Leon:

And you only have so much hardware that you can fit in a building, so you-

Jolie Hales:

And only so many iterations that you’re even able to do.

Ernest de Leon:

Right, so you’re stuck.

Jolie Hales:

So you could even make better graphics.

Ernest de Leon:

Yeah, you’re stuck. I guess the executive producer has to look at it and say, “Look, we basically can use this render farm for about 30 days, and we’ve got 30 scenes we have to render, so you get a date each, so get what you can.”

Jolie Hales:

That sucks.

Ernest de Leon:

And that’s it, you’re done. The limitation is just from a dollar perspective, you can’t, because the budget would just be completely thrown out the water for whatever it is you’re working on.

Jolie Hales:

And that’s how you end up in the uncanny valley.

Ernest de Leon:

Pretty much.

CLIP: Polar Express:

This is the Polar Express.

Rachel Pruitt:

A lot of times we think about HPCs, it might think about the computers or just the hardware in the background, but really, it’s a way of doing things. It’s using a lot of virtual machines, tying them together to do things in parallel processes. There’s an incredible video done by NVIDIA and MythBusters, and they explain the difference between CPUs and GPUs.

Jolie Hales:

Wait, wait, I have to show this to you, Ernest. Okay, so there’s a link here. I’m just sending it to you. So click on this link and explain to us what you see in this video. So for our listeners who want to watch this video, we’ll also include the link in our episode notes. But in the meantime, Ernest, what do you see?

CLIP: NVIDIA/Mythbusters: CPUs vs. GPUs:

All right.

Ernest de Leon:

I see it seems like a robot that’s driving around here.

CLIP: NVIDIA/Mythbusters: CPUs vs. GPUs:

He’s going to paint a picture for you guys.

Ernest de Leon:

Oh, man.

CLIP: NVIDIA/Mythbusters: CPUs vs. GPUs:

In the way that a CPU might do it as a series of discrete actions performed sequentially, one after the other.

Ernest de Leon:

It’s drawing a happy face with paintballs.

Jolie Hales:

So then there’s another robot with an NVIDIA logo.

Ernest de Leon:

On this giant thing.

Jolie Hales:

Uh huh.

CLIP: NVIDIA/Mythbusters: CPUs vs. GPUs:

GPU painting demonstration in 3, 2, 1.

Ernest de Leon:

They’re going to try to land all the paintballs at one time.

Jolie Hales:

Yes.

Ernest de Leon:

The Mona Lisa. Awesome.

Jolie Hales:

Yeah. So the first robot, that’s representing a CPU demo, and it’s basically a robot paintball gun, but it’s a single gun and it shoots one paintball at a time until it makes a smiley face. Whereas the GPU robot, which represents NVIDIA’s GPUs, it’s got multiple paintball guns and it shoots all the paintballs at the same time that makes this more elaborate Mona Lisa, I would say it’s like a 16 bit Super Nintendo Mona Lisa, but still cool.

Rachel Pruitt:

In that video, what they talk about is how you take, maybe, a task, and you split it up into a lot of different tasks. And then what you do is you spread those into different resources.

Jolie Hales:

Like ants working together, doing their solo part while still depending on each other.

Rachel Pruitt:

Think about a house being constructed. You have one person, it’s going to take a little long time, you throw 10 workers at, it’s going to take a lot faster. And now throw a hundred, throw 200, throw 1,000 workers at it, suddenly, it gets done a lot faster.

Jolie Hales:

She admits it’s a simple metaphor for a very technical space, but hopefully that helps those who are new to high performance computing. And in traditional HPC, these supercomputers have existed in-house or physically on the company property of those who are using them to run simulations like computational crash testing. That’s how it was at least before the cloud. But today, more and more of these simulations are done on the cloud or in a data center off company campus, which is where Rachel, of Microsoft Azure, comes in, being a cloud service provider that offers computational power for these kinds of simulations without requiring companies to have these kinds of on-premise hardware systems.

But we’re getting ahead of ourselves. First, let’s go back in time a bit to the pre-pandemic year of 2019. So UD Trucks, at that point, was owned by Volvo, and all of their simulations were run on on-prem supercomputers owned by the company. But then in December of that year, Volvo and Isuzu decided to form a strategic alliance. And as part of that agreement, Volvo would officially transfer UD Trucks to Isuzu in April of 2021. But this created a bit of a challenge for IT at UD Trucks, which was led by Anand, because it meant that UD Truck engineers would no longer have access to the supercomputers they had relied on for these years.

Anand Kumar:

So during summer of 2021, they ask us to develop our own HPC solution and release that computation power, which usually, we are using it on on prem.

Jolie Hales:

They had to come up with their own HPC solution from scratch, and they had to be completely up and running by January 1st, which was just a few short months away.

Ernest de Leon:

In the middle of a pandemic where supply chain was busted.

Jolie Hales:

Right.

Ernest de Leon:

Awesome. Completely doable.

Anand Kumar:

I don’t have a choice. So when you don’t have a choice, you put all your energy in to go and deliver what is needed.

Jolie Hales:

So Anand and his team put their heads together to figure out what options to pursue, should they build a new on-prem system from the ground up? But then-

Anand Kumar:

You have to wait for your hardware delivery, you have to plan your project more efficiently, you have to make a contract, you have to test the solution on prem, you have to wait for racking and stacking, you have to wait for processes [inaudible] be in a spin up, and it’ll take a lot of time.

Jolie Hales:

Time that they did not have a lot of.

Rachel Pruitt:

If you’re on prem, you probably can’t afford to buy the newest thing as soon as it comes out. A lot of times, on-prem data centers are sporting technology that’s, maybe, on average four years old. And so while that’s okay, you’re just going to get higher performance a lot of times moving to the cloud, at least over time.

Jolie Hales:

And when faced with this decision, instead of crawling into a hole or weeping into his Cheerios or something like maybe many of us would’ve done, Anand rose to the challenge and he thought, why not take this short turnaround as an opportunity to scale up quickly, which to him meant, surprise, surprise; cloud.

Anand Kumar:

When you have scalability and agility, cloud is the only solutions we have. With cloud, it will make your life easy. I believe in better life and easy life for ourselves. And with cloud, you can have a immediate test results with better performance without heavy investment and CapEx and OPEX charges and commitment.

Jolie Hales:

But performance was key, and this is one of the fears that we hear sometimes in HPC from engineers, is they’re worried about performance when they move to the cloud. And Anand knew that they could only really justify that move to cloud if they could maintain similar, or better, performance.

Anand Kumar:

And result was amazing.

Jolie Hales:

They started discussions about cloud possibilities in the summer of 2020, and then they chose to work with Microsoft Azure as their cloud service provider, and they developed a plan and they got a contract going by October. But now they have until just January 1st to get this up and running. And by January 1st, they were live, migrated to the cloud.

Ernest de Leon:

It makes perfect sense because if you think about it, even if you couldn’t get the same performance, which I think nowadays it’s very close to similar, the cloud versus on-prem, even if it was, let’s say, 80% of the way there, you can still scale out, horizontally, more and end up getting a result faster on more hardware than you would’ve had physically. So I think no matter how you look at this problem, cloud would’ve solved it regardless.

Jolie Hales:

And I also wonder, in a few short months, could you really even build an on-prem system that was up and running in that amount of time?

Ernest de Leon:

You might be able to get through the environmental impact analysis to lay the foundation for the building that would become the data center. You wouldn’t even get concrete laid on that building in that amount of time.

Jolie Hales:

Yeah, that’s what I would think.

Anand Kumar:

Cloud is not the future, but it’s need of the hour. And if you’d like to scale your environment twice as fast, I think this is the best solution we can have.

Jolie Hales:

And Anand’s words totally jumped out at me because while some IT professionals are cloud friendly, many others, as we’ve mentioned, are resistant for many of the reasons we’ve discussed earlier. But not Anand. He is clearly a cloud advocate, perhaps even a cloud evangelist, you could say. And he says it’s because he witnessed firsthand the speed and scalability that cloud had to offer in the quick timeframe that he had to rebuild their company simulation platform, basically from scratch. So anonymous team decided that a cloud-first strategy was best for them.

Anand Kumar:

In a migration, we wanted to do the apple to apple lift and shift approach or re-hosting approach for typical cargo projects. But when there was an opportunity, we wanted to anchor the cloud first. And why I equal to cloud forces strategy, because it’s easy to scale. Whenever we wanted to add the computer, so I don’t have to add or wait for the hardware order, purchase order, CapEx order, you released lot of administration work. I can just click a button, and my machine is available for me.

Jolie Hales:

He says he got there by breaking up the migration process into smaller steps.

Anand Kumar:

I’m in a better situation now. My applications are live, my users are happy, and my performance is equally, or even better than, on-prem. And then the cost was, I can say that we have a good returns on investment because I don’t have to get the commitment for five years, six years, 10 years down the line, and we get immense vendor support immediately.

Jolie Hales:

But in order to meet the tight deadline and get it up and running in a few short months, UD Trucks turned to, I don’t know if you’ve heard of this company, Ernest, HCL Tech, which is an India-based IT company and a partner of Microsoft Azure, and UD Trucks had actually worked with them previously as part of Volvo. So in this situation, Azure would provide the cloud services, and HCL Tech would help them, basically, execute the migration, so they weren’t in it alone.

Rachel Pruitt:

We sent them to a partner because we have so many great partners and they are really integral to helping our customers get set up in the cloud.

Jolie Hales:

Partners like Rescale.

Ernest de Leon:

Rescale.

Jolie Hales:

Okay, nice.

From supersonic jets to personalized medicine, industry leaders are bringing incredible innovations to market with unprecedented speed and efficiency by using Rescale, high performance computing built for the cloud. Rescale is an easy-to-use computational science and engineering platform that delivers intelligent full stack automation and performance optimization that empowers engineers while giving IT security and control. As a proud sponsor of the Big Compute podcast, Rescale would especially like to say thank you to all the scientists and engineers out there who are working to make a difference for all of us. Rescale; high performance computing built for the cloud. Try it for yourself at rescale.com/BCpodcast.

Rachel Pruitt:

I cannot say enough good things about our specialists and how good they are at HPC and how well they understand the use cases. Anand, they sell from a solutions standpoint, so they’re not going to say for every single customer, this is the product that we want to give you. They focus on what is the problem and how do we solve that. Then they compile a list of products that’s really going to fit those needs and satisfy the customer.

Jolie Hales:

So between all the teams involved at UD Trucks, Microsoft Azure, and HCL Tech, they provisioned Azure resources to create a test environment, and then they began to troubleshoot and optimize the cloud setup, making sure that they could migrate their workloads, their applications and processes without grinding innovation to a halt. And by a miracle, they did it. They made that January 1st deadline.

Anand Kumar:

This was really magic. Every person who was associated with HPC computations using our computation power as a vendor or SaaS provider, or cloud service provider, they help us a lot. They had come with a mindset that let’s solve it together as a one team, as a one project. And that’s how we can see that our computation powers is even better or even equivalent with the on prem.

Jolie Hales:

Everyone came together, including all the IT professionals at UD Trucks, to make the move to cloud.

Anand Kumar:

You need a lot of teamwork, and at UD Trucks, we believe in one team. Even with partnership, we believe in one team, our success and failures are together.

Jolie Hales:

And contrary to others’ fears, they still kept their jobs.

Rachel Pruitt:

You did have to have these people that were serious specialized In all these servers and all this on-premises infrastructure. And while those people are still needed and they can shift into this new role of managing the cloud infrastructures.

Jolie Hales:

When it comes down to it, the migration was pushing full speed ahead, mainly in November and December. So that’s when they were really, really making this all happen. And those are not months that are typically known for high productivity in the tech world.

Anand Kumar:

We had started our project in late November, 2020, and then we had hardly one time including Christmas holiday, so that was very stressful project planning and I can say that, thankfully, everything worked well.

Jolie Hales:

And the last thing Anand wanted to do was to go back to Volvo and ask them for an extension on time to be able to continue utilizing their on-prem systems, which Volvo probably would have granted, but he didn’t want to make that ask. He’s the leader of this team, he wants things to move forward, he’s embracing digital transformation. So thankfully, he actually didn’t have to make that ask.

Anand Kumar:

Thankfully, with good intentions, with good deal and passions and one team approach, we had been successful at the first spin, so we had never looked back.

CLIP:

Celebration.

Jolie Hales:

But it wasn’t without its challenges.

Anand Kumar:

This switch was not immediate. Every stage we failed and we learned from our failure to stand up and we showed the resilience in whatever we did.

Jolie Hales:

While Anand’s goal had been to migrate from on-prem to cloud without their end users sensing any change, apparently, word did get out of what they were doing.

Anand Kumar:

Users somehow got to know that we are moving from on-prem and to the cloud and they were having a different mindset.

Jolie Hales:

When they scaled to cloud, they were using different interactive nodes in different regions.

Anand Kumar:

Initially, when we had shifted the code from on-prem setup up to cloud, the performance was not right. So we did lot of fine tuning at application and also our computations on our end. We did lot of enrollment variable changes, and it took months time to settle down with our user expectations, especially on cloud.

Jolie Hales:

But with the help of all the teams involved pushing forward with the same goal-

Anand Kumar:

I think we had it scaled up quite easily.

Jolie Hales:

One advantage Anand’s team saw in moving to cloud was that they suddenly had access to a wide range of hardware types, and they could optimize performance by picking the hardware that would best work for a certain kind of job. And I know that, for instance, in Rescale’s case, the Compute Recommendation Engine actually uses historical data and AI to even go so far as to make recommendations of which hardware to use for which workloads to maximize performance and cost at the same time, which eliminates the need for a lot of benchmarking, frankly, and AI and HPC is really only going to become more and more powerful, as it is in everything until it takes over our lives.

Rachel Pruitt:

The addition of AI into some of the simulation modeling to, then, further do more predictions, is just incredible and I’m so excited to see where that takes us as we keep going down the road.

Jolie Hales:

Which, oh, Ernest, okay, I need to pause for a second here because I have this really annoying AI-related beef that I need to express, and I’m going to use this podcast to selfishly do that thing.

Ernest de Leon:

Alan Iverson hasn’t been a thing in the NBA for a long time, now, so I’m not sure what your beef is there.

Jolie Hales:

Nice. I barely know who that is.

CLIP: NBA Game:

Wide to shoot for Iverson. Yes. That was for three, his sixth of the game.

Jolie Hales:

So in this age of digital stuff that we live in, everyone pretty much types and they read sans serif fonts, right?

Ernest de Leon:

Which I hate.

Jolie Hales:

So we’re talking, you hate. Yeah. So you’re a serifs font fan, then. So you like more Times New Roman versus Ariel, Helvetica, Roboto. For those of you who don’t know what we’re talking about, which you probably do, but just in case you don’t. So a sans serif font means without serifs. Serifs are the little lines that were traditionally on the tops and bottoms of letters, little horizontal lines. So instead of just having an A that’s like an upside down V with a cross across it, would also have a little two little feet and maybe a hat, if that makes sense. Does that make sense? Just say, “yes.”

Ernest de Leon:

I know what a serif font is. I don’t know if it makes sense, but.

Jolie Hales:

Well anyway, regardless, you can look it up. Google it. Ask ChatGPT, for all I care.

ChatGPT:

San serif fonts are a type of typography characterized by the absence of small decorative lines, known as serifs, at the ends of the strokes in the letters.

Jolie Hales:

The point is, is that all main internet and word processing fonts these days are sans serif. This is a really long explanation leading up to my beef, but whatever, it’s my podcast, I can say it.

Okay, so in the world of sans serif fonts, a capital I and a lowercase L look exactly the same, which is stupid, because now that AI is a real thing and everyone’s talking about ChatGPT, and Sam Altman and all that, my brain cannot make the switch to automatically reading the sans serif letters “AI” as anything other than the name “Al.” And it’s so annoying.

So I grew up with a lot of friends named Al. I’ve worked with people over the years named Al, and for decades, every time I’ve read those squiggles on the screen, they’ve referred to people. And since you can’t teach an old brain new tricks, apparently, at least not my brain, my brain simply does not want to adjust. It naturally wants to see AI as Al. For example, if I type AI into a Google news search, I see headlines like “Al Photoshopping is About to Get Very Easy.”

“Elon Thinks Al Could Become Humanity’s Uber Nanny,”

“Windows Copilot Puts Al in the Middle of Microsoft’s Most Important Software.”

Who is this Al? This Godlike character named Al. When it shows up in email, text threads or something, it’s even worse because it’s perfectly feasible that you might be writing about a person named Al in person to person regular communication.

Okay, well, clearly I have an issue with this because it just seems like everyone on the internet is talking about some random person named Al because somebody in the world decided to make the small L and the capital I look exactly the same in sans serif fonts. And in handwriting, it’s even worse because then you can also throw the number one into the mix to amp up the confusion. And in fact, I think that comedian Brian Regan does a bit about the small L issue.

CLIP: Brian Regan:

Sometimes you’ll get a confirmation code that includes zeros and Os. You write them the same way, they’re completely different keys. They’ll give you the number 1 and the letter I, you write those the same, they’re different keys. A small L looks the same, they don’t care you’ll never be able to communicate this back clearly. I’d like to give a code to these people for their big vacation. Here’s your confirmation code. You’re going to need this for your vacation. Are you ready? Okay, it’s 11I1O00OOI1OO01Il.

Jolie Hales:

Okay. Thank you for going on this journey of annoyance with me, Ernest and listeners, if I don’t cut this out.

Ernest de Leon:

Well, let me tell you, I have hated sans serif fonts for a lot longer time than probably you have. And for me, it all comes back to one thing; passwords.

Jolie Hales:

Oh, totally.

Ernest de Leon:

When you type in passwords, you have this problem.

Jolie Hales:

You’re a cybersecurity guy. It’s like a total issue.

Ernest de Leon:

Complete issue. And so we’ve had to do certain things. For example, by the way, it’s not just serif and sans serif, it’s also things like a capital O versus the number zero.

Jolie Hales:

Oh yeah, the zeros and Os.

Ernest de Leon:

How do you distinguish the two, right? So unfortunately, what has happened is some password generators have taken the tact of, well, I will exclude one of the pair of characters that look the same. So like you mentioned, a capital L or a lowercase l and a capital I. They just won’t use one or the other, which-

Jolie Hales:

Makes it slightly less secure than it could be.

Ernest de Leon:

It reduces entropy, is what we call it.

Jolie Hales:

Yeah.

Ernest de Leon:

What others will do, it’s one of those things where they say, okay, capital letters, we have all in one color, and if there’s a lowercase letter or any of them, we put them in a different color. So white versus blue, and then the O versus zero, they’ll put a slash through the zero.

Jolie Hales:

Yeah, the zero slash.

Ernest de Leon:

To indicate it’s a zero. And so we have to do all of these gymnastics for passwords because of this sans serif issue.

Jolie Hales:

It’s the worst. It’s the dumbest thing.

Ernest de Leon:

It drives me nuts.

Jolie Hales:

You would think that when sans serif fonts were even developed, somebody would’ve looked at that and been a little concerned at, hmm–

Ernest de Leon:

I’m sure someone did, but it was the one engineer in the room that got shouted down because everybody else was like, this looks so much better and blah, blah, blah. And I’m sure there was some other reason, like, perhaps serif fonts incrementally had used up more space on the printed page than the sans serif.

Jolie Hales:

Oh yeah, like more ink or something.

Ernest de Leon:

It was less cost for ink or who knows, right? There’s always some esoteric detail that causes this.

Jolie Hales:

Even if it’s the cost of printing ink, I’m sure that the time that we’ve had to take to decipher whether it’s a small l or a capital I or a number one or something, has been more expensive than had we just spent the dang money to use the Serif fonts.

Ernest de Leon:

Yeah, we should go ask ChatGPT, how much human productivity has been lost as a result of dropping the serifs from fonts. I want it to calculate that.

ChatGPT:

The impact of serif versus sans serif fonts on productivity is not well established and can vary depending on various factors, such as the context, task, and personal preference of individuals.

Jolie Hales:

Back to the important story at hand, here. So, in the case of UD Trucks, which has no capital I or small l in its name, thank goodness.

Ernest de Leon:

Maybe they took that into consideration.

Jolie Hales:

Maybe they did. Maybe they did. So when they had been using their on-prem systems, their simulations and their modeling had performed well, specifically with AMD processors.

Ernest de Leon:

Another non-sufferer of sans serif fonts.

Jolie Hales:

Yes.

Ernest de Leon:

AMD.

Jolie Hales:

True story. And because AMD third gen epic processors also power the Azure HBv3 virtual machine, it made sense to utilize the HBv3, which was eventually upgraded free of charge to the AMD Epyc CPU featuring AMD 3DV cache, which resulted in even better performance than had been seen on prem, which is basically to say, they were able to use even better versions of the same hardware they had been using on-prem when they moved to cloud.

Rachel Pruitt:

All of our virtual machines really are truly optimized for specific workloads. One of the benefits of moving the cloud is you truly just getting the best. And so we have a very close relationship with AMD. We were first to market with that HBv3, it wasn’t even possible to get on-prem for some time after we got it. So when Anand got it, I don’t even think that it was possible to get on-premises. So I think that’s something that he got only from coming to the cloud.

Jolie Hales:

Which brings up another benefit of cloud, in that you don’t have to keep paying to upgrade your on-prem hardware, which is honestly quick to go out of date.

Rachel Pruitt:

Think about how a computer, it gets old after what, couple years? And it doesn’t work as well, the performance just goes down and goes down and goes down. And so when you come to the cloud, though, you can always get the best. You can always upgrade, and that newest technology always performs so much better.

Jolie Hales:

And for UD Trucks, once the dust of change had settled, the engineers who had been understandably skeptical of the plan are doing well.

Anand Kumar:

They’re feeling good. Ask her, initially this felt that nobody had done this, and then they’re feeling proud, now that we are able to scale up the same performance and the same functionality in the cloud, and they are super happy.

Jolie Hales:

Which makes Anand happy.

Anand Kumar:

If user is happy, we can produce more trucks and we can get more number of sales. If we get a more number of sales, our revenue will be also impacted.

Jolie Hales:

And looking back, Anand feels like he made the right call.

Anand Kumar:

We can scale up our environment as we want. We can add the port, we can add the node in the click of the buttons.

Rachel Pruitt:

Microsoft’s mission is to empower every person, every organization on the planet to achieve more. So I think a lot of times that hesitancy is the unknown and being worried that things will go wrong or it won’t work as well as on-prem and would encourage those people to just consider Microsoft as a partner. And then a lot of times, you bring in another partner, so you’re not doing this alone. You have a team backing you that can help you along the way and help you get everything set up.

As Anand said, yes, there are challenges. There’s a lot of learnings that you make along the way, and it’s really not a straight comparison of on-prem versus the cloud as being mentioned. There is changes that have to be made to make that infrastructure work really efficiently on the cloud. But again, that’s what your Microsoft experts are there for, that’s what your partners are there for, to help you along the way as you do that.

Jolie Hales:

And not every case is just like UD Trucks where they had a couple short months to completely redo their HPC situation from scratch.

Rachel Pruitt:

Some people move fully over to the cloud and it’s whatever works for you, and just knowing that Microsoft and our partners are here to actually help you through that process.

Jolie Hales:

And for Anand, he’s proud to say that UD Trucks is one of the first manufacturing companies in all of South Asia that has successfully migrated to the cloud, along with all their applications, which really does make them trailblazers in this way.

Anand Kumar:

We were the one who had decided that we can go with the cloud with high performance computing and we are there.

Rachel Pruitt:

This is, by far, my favorite part of the job is just seeing how all these businesses implement HPC, and they’re changing the world. How we do things is changing every day at such a rapid pace and it’s incredible how things are being manufactured, to curing drugs, to forecasting weather. It’s just a really, really exciting space.

Anand Kumar:

You create a region and try to achieve it and it’ll create a better life for you, better life for the people and better life for the planet.

Jolie Hales:

A better life. Isn’t that what computational engineering is all about, in the end? Making life better for all of us? Kumbaya, my Lord, kumbaya.

Ernest de Leon:

Yeah, but one of the things I would like to point out, I would actually challenge people and say, look, if you’ve got a project coming up that is HPC related and you are looking at the options between building an on-prem solution, putting a new cluster out, whatever the case is, and/or using the cloud, what I would say is, using the cloud is not a long-term commitment. It doesn’t have to be, I should say. Buying an on-prem cluster and housing and operating that thing is a long-term commitment. So what I would say is do this; have two competing projects that launch on the same day.

Jolie Hales:

Oh, yeah.

Ernest de Leon:

See which one gets to their solution first or do it the other way, go cloud first because you know you can do it faster, and then throw it out there and see what happens. If you have a revolt, fine, then you go the other way. But if it works, you’re now in a system where you decide what the spend is, you don’t have the long-term commitment, and you don’t have to manage all that stuff, it’s out of your hair entirely. That’s exactly what I would recommend here. This is the Coke versus Pepsi challenge. Try it, put your users on there and see if they complain about it.

Like Anand said, you’re getting better and more advanced hardware all the time in the cloud. When you put a new cluster into use, you are stuck. That is a snapshot in time. That hardware, the minute you power that thing on is obsolete. That is not the case for the cloud. So I would say try it, if you don’t like it, if your users don’t like it, just back out, you’re done. But I have a feeling that, that’s not going to be the outcome, that you’re actually going to find your users are happier doing it that way.

Jolie Hales:

In fact, do this, people, listeners, please do this and then reach out to us on bigcompute.org and we’ll make a podcast out of it.

Ernest de Leon:

You might be surprised what the outcome is.

Jolie Hales:

Ernest. You’re not just a cybersecurity guy, you’re a marketing genius.

Ernest de Leon:

That was in a former life. I now just look at passwords all day.

Jolie Hales:

With small ls and capital Is.

Ernest de Leon:

Drives me nuts.

Jolie Hales:

Well, we’re in the same boat.

Ernest de Leon:

Oh, I can’t even tell you. It’s been years that I’ve been trying to fight this fight, and it doesn’t work because you get overruled. Or even 90 plus percent of people don’t care. They can read it, right? It’s not like they can’t, but those of us that need absolute clarity, we struggle with this and it’s a private struggle that we all internalize. We bear the burden alone.

Jolie Hales:

Until we externalize it on a podcast episode.

Ernest de Leon:

On a podcast, right.

Jolie Hales:

Well, maybe quantum, once that gets up and running, the password thing will be inept and we’ll just have to deal with Al.

Ernest de Leon:

Yeah, something. But yeah, I have complained about– just like I’ve complained about the caps lock key for at least 20 years. What is the point of that thing?

Jolie Hales:

Just hold down shift.

Ernest de Leon:

We don’t have typewriters anymore. I understand it took you a significant amount of effort with a typewriter to push a key down, right?

Jolie Hales:

Yes.

Ernest de Leon:

Or to hold shift and push a button, but now you’re on a keyboard that it just doesn’t make sense to me. Some of these holdovers.

Jolie Hales:

And have shift on both sides of the keyboard. It’s not like you can’t access the shift key. You’ve got double the opportunity to hold down shift.

Ernest de Leon:

Yeah, caps lock makes no sense. Now, again, there’s probably somebody out there who’s like, “actually, I use a mainframe to do something and I have to do everything in capital letters” and, okay, fine. You are the person who needs the caps lock, right? But for the majority of people, since the end of the typewriter, a caps lock has served no purpose and they won’t remove it from the keyboard, it’s like-

Jolie Hales:

No, it’s going to be around forever.

Ernest de Leon:

They’re insistent that it’s going to be there until the thing goes down.

CLIP:

Yeah, the caps lock just won’t come off, right?

Jolie Hales:

Yes. Well, Ernest, it’s been too long since you and I have recorded our gripes about technology, so I’m glad that we got that out of our system.

Ernest de Leon:

We need a Festivus podcast. Airing of the grievances.

Jolie Hales:

That’s true.

CLIP: Seinfeld:

Happy Festivus, everyone.

Jolie Hales:

Special thanks to Anand and Rachel for joining us on this podcast and for also tolerating all the extra stuff that we put into this episode that doesn’t have to do with the awesomeness of Microsoft Azure and UD Trucks. Yes, this is the way we roll here, at Big Compute. And of course, to learn more about UD Trucks and the cloud HPC services Microsoft Azure has to offer, you can head on over to the episode notes page on bigcompute.org where you can find links to the topics we’ve covered today, including a link to Rachel’s blog where she actually discusses what high performance computing is for the layperson. There’s also a link to a case study about this very story that we’ve been talking about today.

Ernest de Leon:

You can also support us by leaving a rating or review on Apple Podcasts, or better yet, telling a friend about us.

Jolie Hales:

That’s right. And with that, thank you for joining us today.

Ernest de Leon:

And remember everyone, always use multifactor authentication and 3, 2, 1 backups.

Jolie Hales:

Stay safe out there. I’m trying to think of a joke to say about Al at the end, but I can’t think of anything. I think Al just is a joke. I’d be mad if I– it’s like the friends who are named Alexa, right?

Ernest de Leon:

Yeah. My uncle and my cousin and son are both named Al.

Jolie Hales:

Yeah, you should ask them how they feel about this.

Ernest de Leon:

Yeah.

Jolie Hales:

AI Al, this is serious issue for them.