HPC 2.0: How CIQ Is Simplifying Supercomputing for the AI Era

>> I can make it real easy. a delighted user gets to actually do their job. So in the HPC space, like all things, there's management, there's patching, there's building, there's deploying, there's all of this that has to be done. And we have a viewpoint of let users work, let them do the things that they want to do and not have to spend their time managing that infrastructure. And that's our approach to the operating system. That's our approach to our cluster manager. And that's our approach to fuzzball. Give the user the ability to do their work. Any other thoughts on that? >> Yeah, absolutely. I mean, um, in the HPC space, we have, and I say we like admins and staff scientists within HPC have traditionally thought to ourselves, well, these users, they need to learn more.

We need to make these users computer scientists. They need to learn how to use the batch scheduling system and Linux and all this other stuff. And if they don't, well, then that's a shortcoming on their side. I have worked with a team and the CIQ team um is a team that says no that's wrong. These users they already they know their domains and their domains are complicated and they need to know um a lot about their science or their expertise whatever that is. We don't need to make them computer scientists too. We're failing if we're doing that. What we need to do is we need to let them do their science and just give them the power to do it without having to learn all these new tools.

So, I'll tell you what I've been impressed by, and it's interesting. Just a moment ago, I met one of the newest members of your team. And I said, 'What's your background? And he said, 'Well, media, where I've been dealing with the orchestration of these things, working for big television networks, and it just maps directly with what with everything I've seen with CIQ. People who have actually dealt with the real world problems, how to address them, now coming to the other side of the table, >> if you will, to to to solve those problems. I mean, does it feel like that to you? >> No, 100%. So, I mean, for the last before this, I was in the storage industry.

Um, similar thing, helping customers solve their their media problems, solve their backup problems. Uh, and from the same concept here, it's just at a at a different piece of it. So, how do we help them actually take a step beyond of either a giving them a more compliant environment to operate in? Especially in the in the government space, compliance matters, security matters. How can we give them a more secure operating system and a secure environment to run inside of? And then finally, performance. And depending on who's doing what and where it matters, those three kind of rotate as a what's uh what's most important and what falls second and third on that list.

>> So when I talk to people in the HPC world, they talk about three things. Compliance, security, and performance issues they need solved. What do you do to help them with those three things? Yeah. So we take the baseline uh Rocky Linux operating system that CIQ now delivers in a product called Rocky Linux from CIQ uh and we create a variant called hardened which we refer to as RLC Pro Hardened or Rocky Linux from CIQ hardened. And what comes along with that is first off we we preh harden everything meaning that we do code level changes and introduce certain things like the Linux kernel runtime guard which protects against if somebody gets behind the castle walls if you will it limits the blast radius of what type of destruction they can create uh and access to your systems.

Uh and then beyond that we will take anible playbooks uh to make it easier to do uh compliance deployments uh so achieving uh CIS benchmarks uh applying diesig to these environments and these are really important to our government customers uh and clearance uh focused users uh and then finally beyond that we will do kernel tweaks and user space tweaks uh that allow for a more performant deployment depending on where the customer is needing to run their operating system and what the work workload may look like. Yeah, if I could just add on, I mean really it's it's all about just recognizing that customers are different and some customers, you know, just really want to focus more on stability and some customers really want to focus more on things like performance and so you can give the customers, you know, that choice.

Some customers might be in a situation in which they have to focus on security. So you give the customers the choice and you you know you take them out of the same box and you put them in you know in separate places where they can choose. >> Okay. So lightning round take you guys by surprise. You step into an elevator and you have 15 to 30 seconds. You just you learn that this person is someone in charge of a complex HPC environment and you grab them by the shoulder and you say dude or dudette. just remember these two or three things to get them to call CIQ.

What are those things that you're going to try to impress upon them? Yeah, that short elevator ride >> that we're going to abstract away not only from the the users perspective, but also from the admin perspective a lot of the stuff that you're working on right now. We're going to make things a lot easier for you and for your users that you get fewer tickets as well. Okay. Now, you get in on a different floor and it's the same person. What else do you tell them? stop worrying about the infrastructure uh as a whole and let us handle that from top to bottom. Let us make it compliant.

Let us make it secure and let your users actually get to work. Wasted time on their part managing environment is wasted resources and wasted cost and that's not something that anybody has the ability to to withstand. >> Fantastic. Well, thanks for joining us here at CIQ Town in the Dave Lounge with the three Daves. This has been a take five from 65 Media. Thanks for joining us and stay tuned for more interesting content.