In Red Hat’s new podcast, Ryan Kraus, Staff Data Science Solutions Architect at Red Hat, discusses trends in edge computing across the public sector and Red Hat’s collaborative AI-driven solution that enables high performance data processing in the field.
Rich Savage: Thank you for joining us today. This is Rich Savage with the Red Hat team at Carahsoft, and I'm here with Ryan Kraus, Staff Data Science Solutions Architect at Red Hat, to talk about trends and edge computing across the public sector.
Ryan Kraus: Hey, how's it going?
Rich Savage: Great, great. Glad to have you here, Ryan. To get us started, can you please take a moment to provide our audience with a brief introduction of the world of edge computing?
Ryan Kraus: Yeah, sure. Edge computing is probably one of the broadest buzzwords I think that we have right now. There's really a whole spectrum that falls into the category of edge computing, and all of it boils down to simply wanting to put compute closer to where the action is. We like to think of extreme edge cases as maybe some IOT devices that are gathering sensor data and feeding them back in, or maybe we'll have a remote office and we want some systems there to act as sort of some edge compute power. Or, maybe it's a cell phone tower that has a mini data center at the base of it, providing some compute for the tower. So, really we have this broad spectrum there that everyone kind of falls into. But like I said, it's mostly getting compute close to the action.
Rich Savage: That's great, and yeah, absolutely I agree. We're hearing it being used in many different use cases. I guess maybe let's hone in on that a little bit. What are some of the primary IT challenges that public sector organizations face that edge computing can really help to overcome?
Ryan Kraus: Especially in public sector, data security is always tantamount to basically everything. And one really nice thing about edge computing is that now, instead of having to send sensitive data over the wire, the field, back to my data center, I can actually just process it locally in the field and keep that sensitive data where it exists instead of exposing myself to the risks of distributing that out.
Rich Savage: I always like to ask the question, why now? Right? Why is edge computing more relevant now than ever, and why is it rapidly evolving at such a high pace in the public sector?
Ryan Kraus: I think that's always a valid question with every technology, is this the appropriate time to jump in and start looking at it? I think with edge, one of the big reasons why it's taken off now is we finally have reasonable amounts of compute in a small enough footprint to do something worthwhile.
Another compelling argument there is that Kubernetes has really given us this consistent layer to be able to run and manage our applications. A lot of that burden of deploying these systems and then managing them once they're in the field has really been obfuscated by Kubernetes. So, we're kind of at this unique turning point where the hardware and the software are really catching up with what we really want to do.
Rich Savage: Yeah. That term, "catch up," right? We need that technology to move these new initiatives forward. You hit on a couple of them, I think certainly on the Kubernetes side. Are there any other systems or technologies that government agencies need to have in place, or need to modernize in order to really take advantage of edge computing solutions?
Ryan Kraus: Yeah, that's a good question. You can't really move out to the edge unless you have some kind of strategy for getting your software out there. We mentioned Kubernetes, but that's really what containers are. It's a portable unit of work that I can easily ship out to remote systems or even disconnected systems. Really being able to run those in your data center is pretty key to being able to then branch out to edge locations, because the edge just becomes sort of an extension of your operations in the data center. So, it's really beneficial to have a consistent developer and operator experience both in your data center and then out to the edge. It kind of bridges that gap.
Rich Savage: Thank you, Ryan. I'm going to steal a quote from your president and CEO, Paul Cormier. He said that, "If edge computing is going to be a realistic future for enterprise IT, it needs the hybrid cloud and open source to thrive." Why is open source the ideal foundation for operating a comprehensive edge-computing solution?
Ryan Kraus: Yeah. One of the things that's fun about source is it gives you a lot of power to mix and match, build out a solution that fits for you. A lot of times with proprietary software, there is a way that it's meant to work, and you have to use it that way or go somewhere else. The great thing with open source software is if there's a way that something works that ... Let's pick on Red Hat a bit and say that there's a piece in OpenShift that, maybe I don't want to use Jenkins as my CI/CD pipeline. Everything is open source. I can drop in a new pipeline management tool, or I can swap out pieces here and there to build a custom solution for me.
And when we're talking about going out to the edge, everything becomes very hypersensitive, that if I change something back in my data center, I now have to change it at potentially a thousand edge locations. So, having that ability to really architect a solution that's right for me and my enterprise becomes especially important with edge computing.
Rich Savage: We know that Red Hat is all-in on the open hybrid cloud strategy and vision. What role does edge computing play in coming together with that hybrid cloud vision from Red Hat?
Ryan Kraus: My initial role at Red Hat was actually as a cloud solutions architect, and my background has been in scientific computing, so I transitioned over into focusing on data science. It's a very common question where, how is something related if I'm running it in a battlefield or if I'm running it in some nebulous data center that I don't actually have access to?
But in reality, any edge computing device, you're going to be streaming some amount of data back home. Be that your data center, be that AWS', or Google's, or Microsoft's, or IBM. Whomever owns the data center, there's going to be some amount of connectivity back in most situations. And being able to pick and choose where you send that data back to is crucial. Or sometimes, even if you send it at all.
One thing that's really compelling about OpenShift and the solutions that we're building around edge computing is, you can gather your data locally on the edge, and then send it back to any private cloud or public cloud, or none at all. It's all really up to you. You're not hamstrung into making sure your data only goes in one direction. Being able to have that flexibility and choice on how you come back from the edge is crucial.
Rich Savage: And moving to the edge seems like a no-brainer for especially our state and local and education customers. Can you please provide a few common roadblocks to pushing workloads to the edge?
Ryan Kraus: It's a good question. I think right now, most of our customers have just started evaluating this. So, there isn't a fully fledged-out blueprint on how they want to do this yet. It's not like, if I'm going to stand up a new data center, then I know exactly what I need. I could build on a BOM for a data center pretty quickly, but building out a bill of materials for a new edge site is a little bit more difficult just because we haven't done it before.
A lot of our customers that we've gotten feedback from have said that really what they want on the edge is an appliance-type device, something that they can just pick up off the shelf, drop in, and then have most of their decisions solved for them. Really, that's what we are looking to do at Red Hat, and what we're doing right now is trying to make people more productive on day one by just giving them that opinionated architecture of what an edge computing site looks like for a particular range of use cases, and helping them drop that into their current environment.
Rich Savage: That's a great segue into this new solution I've been hearing a lot about from Red Hat, HPE, and Nvidia. I believe they're calling it KubeFrame for AI-Edge. Is this the example of what you're pointing to of, hey, here's a pre-configured appliance ready to go for a customer?
Ryan Kraus: Yeah. So, KubeFrame has been basically my life for the past couple of months now. It's been a really interesting project. HPE supplies the hardware, Nvidia supplies the GPUs, and Red Hat and supplies the software that brings everything together. So essentially what KubeFrame is, is an opinionated architecture for edge deployments. There's a certain level of redundancy built in, a certain level of capacity built in that makes it appropriate for a particular range of edge use cases. What I really like to refer to it as is, it's basically an entire data center in less than a cubic foot.
Rich Savage: I guess, what specific technologies or software from Red Hat is integrated into this bundle that makes everything work together and run together, as you said?
Ryan Kraus: Obviously the bulk of the platform is OpenShift. OpenShift is what manages the workloads and gets everything running. For any listeners that aren't aware, OpenShift is Red Hat's distribution of Kubernetes. And then the other major piece of software we have that I think adds some really cool capability here is what we call OpenShift Container Storage, OpenShift Container Storage is built off of Ceph, and it provides object file and block storage to the cluster.
What's really interesting is that since you have object storage now on this cluster, you have S3 buckets that are local on the cluster that you can then sync back to Amazon, or Google, or Microsoft, or back to an S3 bucket you're hosting on-prem. So, it really gives you this kind of remote cloud feel without being tied into a particular vendor.
In addition to the Kubernetes platform and the container storage that we put on this box, we also have bundled in something that we call Red Hat Advanced Cluster Manager, which basically is a way to manage all of these remote clusters from one central location. So, ACM can manage security policies on your clusters, they can manage the workloads on your clusters, and it can distribute the software out to your clusters as you need it.
So, we think as edge computing continues to grow, this is one of those things where if you don't plan for it accordingly upfront, you can easily acquire a lot of technical debt on how you're going to manage these boxes. We've gone ahead and bundled in that software. Finally, all of Red Hat's Marketplace is available on these edge devices as well. So, any third-party software that Red Hat partners with is available to consume as well.
Rich Savage: Thank you for that. There's a lot of stats out there that I read all the time. One of them being, 75% of data will live at the edge by 2025. Another one I'm reading is that IT organizations will be spending about 30% of their budget on edge cloud computing over the next three years. And for our listeners, if they want to learn more about edge computing and the role that KubeFrame for AI-Edge, a solution from Red Hat, HPE, and Nvidia, where do you suggest they go to find out more about this solution?
Ryan Kraus: Right now we're offering KubeFrame specifically through Carahsoft as a bundled offering with all the HPE, Red Hat, and Nvidia bits under a single skew, just to make deployment and purchasing a little bit more streamlined. So, reaching out to your HPE or Red Hat rep or Carahsoft is probably the best way to go. I also do believe Carahsoft has a landing page with more information.
Rich Savage: Perfect. Yeah, absolutely, that's great. Ryan, thank you again for your time, all the insights, and information. We really appreciate your expertise coming from the field and working with these customers directly. So, thanks again for your time.
Ryan Kraus: Yeah, thanks for having me. It was a lot of fun.
Rich Savage: Of course. And for our listeners, please feel free to visit the website and resources that are on our landing page. Of course, at any time you're always welcome to reach out to redhat@carahsoft.com, or (877) 742-8468. Thank you for your time. I hope you were able to get some valuable information from this podcast, and we'll see you next time.