Getting the Scoop on HPC-as-a-Service With DUG at SC24

Bill Panos: Hi, I'm Bill [Panos, Sr Program Marketing Engineering Manager] with Solidigm and I'm here at Super Compute 24. I'm here with Ron [Schop, Executive Vice President] from DUG. All right, nice to meet you, Ron. Can you maybe talk a little bit about your company? I understand your organization is HPC-as-a-service [high performance computing-as-a-service]. What's that about?

Ron Schop: Yeah, so that's one of our product lines. DUG was founded as a geoscience data processing company. We handle huge amount of data, so for that we had to start some very innovative solutions to keep our power bill down. That is our liquid cooling. So we came into the liquid cooling game about 15 years ago, we started it already. Probably since 10 years [ago] we [have] run all our compute on liquid cooling.

Bill: Got it.

Ron: Using our liquid cooling, this is actually our latest little innovative launch, which is the Nomad product line which is a data center contained in a sea container.

Bill: Got it.

Ron: It's just a nice example of how we're always trying to push the frontier of innovation. So we process large data sets, but we always have spare compute, of course, between the projects. So we're selling that as HPC-as-a-service.  We're a bit different than the others, again, because all our HPC-as-a-service is based on liquid cooled solutions.

 Bill: Right.

Ron: Less power, less water, cheaper. So we're able to sell HPC cycles a bit cheaper than the others as well.

Bill: Could you tell us a little bit more about why a customer would tap a solution like this as opposed to just simply building an on prem [on premises]?

Ron:  Good question. I think it gives our customers a real nice peace of mind that they're doing right with the earth by using a sustainable solution like this. It's a lot cheaper using bare metal machines, which basically becomes their own machine but just on our premises [rather] than building it on their premises. It's not too hard to set up a liquid cooling, but as you might see behind you, which is our data center in Houston, once you start building at a capacity it takes a bit of space and a bit of weight. So why not use our HPC cycles. It's as easy as it gets.

Bill: Awesome. Well you know with the the advent of AI becoming more important how has this change your profile and what you're doing as it becomes more important with AI and HPC type workloads.

Ron:  Yeah, correct. So we actually see a swing from HPC as a service to AI/ML-as-a-service.

Bill: Oh really? Okay.

Ron:  Which you know HPC is probably more memory where, you know, AI is more compute. But the biggest thing, of course, is everything is getting bigger. Everything is getting hotter. Everything is using more power. So the difference there is that we need to start creating more and more heat rejection for our emersion cooling tanks. So we're working on improving our FCMs, our fluid control modules. We actually, last month, we did some tests with some new modules that doubles our cool heat rejection on our tanks.

Bill: Oh wow!

Ron: So that is great progress because AI/ML, you can translate that by everything needing to be bigger, including your storage needs of course.

Bill: So the storage needs are getting bigger. So you could necessarily take advantage possibly of Solidigm's larger capacity drives of which I have an example here of our 122TB [SSD]. Do you think this would be helpful in the future?

Ron: They're awesome. Yes, now we would like to get our hands on that ASAP.

Bill: All right awesome. Well, maybe finally, Ron, you can tell us a little bit about where folks can get more information about DUG and [how] they can contact your team.

Ron: Yeah, sure. Come to our website, dug.com, and you can go from there.

Bill: Okay, all right, well thanks very much.

Ron: Not a problem.