Transcript
But before I do that, I'm just going to do a little, you know, sort of vocabulary thing here. Jeff mentioned in his presentation, you know, he's talking about the Gartner Magic Quadrant. The edge computing or the, you know, edge computing is not just about location, right? It also is about feature set. We just saw a new name for running things on-prem. That feature set is essentially what sc-platform does. And I'm sure almost everyone in the audience is either running sc-platform or has customers that are running sc-platform. So I do want to talk about the location a little bit. From my perspective, if you're running R clusters somewhere without razor wire, backup generators, and security guard, you're at the edge. There are obvious edge deployments, like the 200-plus container ships at sea we have or the In-N-Out Burger food trucks that drive around with a cluster on them. And we're deployed at retail edge like Delhaize and Royal Farms, who you're going to hear from tomorrow morning. If you're an MSP and you're deploying scale computing at customer sites, you're doing edge computing. You have scale clusters at customers that you're managing remotely. About half of our top 20 fleet manager deployments are MSPs, like KT Connections, IT Guard, and RBS. The mid-market companies are the customers that Jeff and I and Phil and Collier and Nate were thinking about when we started this company. And you're right in the middle of this edge computing landscape. When the entire world is saying, everything's moving to the cloud, we were like, come on, that's stupid. Not everything can be in the cloud. So I understand that many of you don't think of yourselves as edge computing, but I promise you, you are. And the reason I wanted to talk about that is because I wanted to show you our product vision to make building, deploying, and maintaining applications at the edge as easy as it is in the cloud. And that's what we're trying to do for all of you. We want to give you the hardware and software infrastructure you need to make building, deploying, and maintaining the applications your organization cares about where it needs them. And make that easier and better than it's ever been possible before. So I'm going to welcome up two of my senior product managers, Mitch Etteson and Taylor Light. We're going to do something a little bit different today, hopefully it's fun. We are going to present our product roadmap a little differently. We're going to talk about new and upcoming features through the stories of two companies. Some of these features are available today. Some will come soon, and some will come in the future. As we're going through these stories, we'll talk about the features like they're here today, as if these guys are deploying them and using them now, and then kind of I'll cut in and tell you when they're coming and that sort of thing. So first, Mitch is going to be acting as the IT manager of a mid-market manufacturing company. Mitch, tell us about your company. Yep, absolutely. So as Scott said, some of you may know me as Mitch Etcheson, the product manager. But today, I am Mitch Etcheson, the director of IT at Etcheson and Sons. No nepotism in my role, I assure you. And we specialize in making modular storage systems. So we are a small manufacturing organization based in Batesville, Indiana, with three total sites, two manufacturing locations, and then one home office, all running on-premise infrastructure And due to some changes in the hypervisor market, we're looking at making a change. I have four total IT admins, so it's a pretty small shop, and we manage everything in-house. Cool. Taylor will be playing the part of an IT manager for a nationwide retail chain. Taylor, can you tell us about your company? Yeah, I'm Taylor Leick. I'm the senior infrastructure engineer at Faraway Pharmacies. We are a nationwide pharmacy chain that is headquartered in North Dakota, and our locations are spread all across, coast-to-coast. We have about 1,000 locations, 50 of those are corporate, where the rest are franchisees. And so we've got about two dozen IT staff up in North Dakota, but we're doing everything. We're doing help desk, we're helping Jane with her Windows laptop. And so really, the in-store infrastructure is primarily managed by just three people. And one of the things you mentioned to me when you were talking about this is you really kind of think of yourself as an MSP for your franchisees. Yeah, absolutely. So these stores are open. If their IT is not up, they are not doing business. So we're on the hook for that. And yeah, we are definitely a service operation, and they are our customers. All right, so there's been a big shakeup in the industry. The number one hypervisor in the market was acquired by this huge company, and it's impacted both of you guys. Mitch, can you tell us what's happening on your end of things? Yeah, absolutely. So we are a long-time user of the now BadCom hypervisor, and as soon as that acquisition went down, we noticed very quickly things like support knowledge-based articles were getting lost in the shuffle as they were changing from the previous company's support portal to BadCom support portal. In addition, just getting support text on the phone was increasingly hard, so our support experience was going downhill pretty fast. Of course, all of this was happening against the backdrop of weed or renewal coming up, and it was really hard to even just get the quotes to know how much we needed to pay for the renewals. And unfortunately, when we finally did get our hands on those quotes, they were much higher than we expected or had ever seen in the past. And so for all of those reasons, we have been evaluating a change to a different hypervisor solution. Right. Taylor? Yeah, everything Mitch said is true. BadCom essentially said, hey, screw you, we're not friends anymore, don't call me. They actually eliminated our license type. After we deployed in a thousand stores, we made a major investment, and it really shook us up for a second. And so as a retail organization, every dollar counts. So when I try to go to my boss and say, hey, licensing just went up 50%, 100%, there's no room for that. There's no room for that in the margin. And so we had to scramble pretty fast before our renewal was up. All right. So you've both decided to move to scale computing. Congratulations. Taylor, you have to deploy a thousand locations, a thousand clusters across the country. Deploying that many clusters must have been a lot of work. Can you tell us about that? It was a lot of work. Honestly, in a thousand sites in a year, this might sound like a fictional company scenario, but we knew it was possible because of rural farms doing the same pace, actually, and as Scott said, you'll hear from them tomorrow. So in order to make this possible, it basically worked out to deploying four hyperconverged clusters to site every weekday for the year. So that meant there was absolutely no allowance for friction, no additional steps allowed. We had to ship these things out to the site knowing that when they're unboxed, they're gonna just work. So in order to do that, we went with the NUC platform, and in our staging warehouse, we ended up using the new and improved zero-touch provisioning. So zero-touch provisioning's been part of scale computing's fleet manager product for a couple years now. But with the new improvements coming, and that we were able to leverage in our deployment, when you're staging hundreds of clusters, you end up, you know, you fat-finger a few things after lunch. And so we go through, everything looks like it's flying, but, oh, whoops, I'm gonna hit an error, because, okay, so now, instead of just erroring out and, you know, now I gotta get my keyboard and mouse out and resolve this, I can see the error, and it's pointing me directly to what's wrong, which is the second and third nodes can't communicate to the first node, that's a backplane issue, I hit edit, I update my backplane VLAN ID, which was set to the default accidentally, and as soon as I do that, fleet manager resends the new config, proceeds, and we're through. So no need to resort to the manual method, it's all handled in the fleet manager. Awesome. So, we've done a lot of work on node lifecycle management, so this includes things like add and delete, and, you know, we spent over, more than the last year working on it, and without this, there's no way Taylor could have met his timeline, and he would have got some pretty nasty cease and desist letters from BadCom. So node lifecycle management is coming as part of 9.6, we're gonna initially, the HyperCore piece of that is gonna be launched this quarter, there'll be early access to that shortly after platform, and I'll talk about that a little bit later, and then the fleet manager piece will be available in Q4. So Mitch, you only deployed three clusters, but this wasn't a run-of-the-mill deployment either, can you tell us about it? Yeah, absolutely. So, even though we had three sites versus Taylor's 1,000, we still were looking forward to the prospect of using ZTP for the initial cluster deployment, if anything, just being able to do an app fleet manager in my office versus, you know, on the ground, in the server closets with a keyboard and mouse, it was a much more pleasant experience. So we did use ZTP, but we had some unique hardware requirements, and Scaled did not offer a, like, fully pre-built certified appliance with everything that we needed, we're a long-time XClarity user for our out-of-band management needs, and as such, we really wanted to use Lenovo, but again, had some specific configurations we wanted, fortunately, we have a long-standing relationship with a channel partner who is also a scale-authorized imaging partner, meaning that we could work with them, get the exact config we wanted, and then they image the nodes with HyperCore and then ultimately pass those over to us for our deployment. So we got the ease of use of ZTP with fleet manager while also checking off all those requirements for the hardware side. Awesome. So partners have been telling us for years that they'd love to source their own hardware. They have brand loyalty and real dollars tied up in rebates and marketing development funds. So we have full coverage to similar builds as the appliance right now available on Supermicro, Dell, and Lenovo. There are some limitations on the small form factor stuff, but most of the main stuff, or all of the main stuff, and then for partners in the audience, you can use this QR code to sign up for the wait list for this program. We are running. This program is happening now. We've got a handful of partners, and it's growing. And we have a sneak peek of this tomorrow at 2.35 with Scott Mann and Craig Tyriac, so if you're interested in that, be sure to check it out. Okay. So it sounds like we have the ability to deploy clusters pretty easily, but that doesn't really do much. You need to get the applications on there. So Mitch, how did you migrate your applications from BadCom to scale? Well, so in addition to being a long-time BadCom customer, we're also a long-time Veeam user. We've really relied on them for years now for all of our data protection needs, and we really love all of the features of Veeam. So it was something we wanted to continue using with our new infrastructure platform. Fortunately, scale computing also has an integration just like BadCom, and so we're able to just restore our existing backups directly on our new HyperCore clusters, and then from there, set up Veeam to continue backing up those applications from then on. But the migration was really easy because we got to use existing tools that we were already comfortable with, and then continue to leverage all of the Veeam features into the future. That's awesome. Obviously, this is one of the most exciting things happening at platform this week, the announcement of Veeam. But like Mitch said, it's not just for migration. It's full backup support, agent-less support directly on HyperCore. Veeam's going to be launching this in their 12.3.3 release and their 13.0.1 release. So we're expecting full production support by the end of the year. Veeam has a talk on this. Ron Martin and Tim Hudson are going to be talking tomorrow afternoon at 1. I'm doing some demos and all kinds of cool stuff, so be sure to check that out. Oh, they also have a booth, so be sure to hit that, too. Taylor, you need to deploy your applications to 1,000 clusters. Can you tell us about how that went? Yeah, so we had them all sitting in the warehouse, but they're just clusters, not going to do much for our stores, right? So we knew that Scale Computing had an Ansible collection, but we're a Terraform shop, so we were really excited that they had just launched a Terraform provider. We have a lot of investment in that tool set, and so we were able to use Terraform to basically leverage the full power of the HyperCore API really easily. So we were able to import our media, boot up those applications, and even do all the validation of the applications while they were still in the warehouse. And as a final step, we used Terraform to re-IP all the nodes to match the store configuration so we could box them up, ship them out, and have that really low return rate that we needed to make our massive deployment a success. Awesome. So basically, you got the cluster up and running through ZTP, then pushed the Terraform go, and it did everything, shut it down, and shipped it out. So really low-touch kind of deployment situation. Absolutely. Yeah. So Terraform is available now. It basically supports everything that is in the REST API, and anything you can do in the UI you can do in the REST API. So basically, whatever you can do with the cluster is available in Terraform. And you can get it on Scale Computing's GitHub profile. There's a QR code there. Okay. So you're both up and running. Mitch, can you tell us about how Scale helps you with your day-to-day management tasks? Yeah. Absolutely. So I'll do you one better. I'll do a quick demo on what daily management looks like for us. All right. So it kind of starts in Fleet Manager. That's where I begin my daily activities. I get a view of my clusters, and I notice through the Fleet Manager metrics tab if there are anything that looks wonky in my deployment. And sure enough, I have an application here that is consuming much more storage than I would expect. So I go to the metrics tab in Fleet Manager, which provides not only cluster-level metrics, but also application-level metrics. And so I drill down into this specific cluster and see, yeah, sure enough, there is a virtual machine that is consuming more storage than I would expect. So I want to go and dig around, see what's happening, fix that problem. I use Secure Link to navigate directly to the cluster without needing to be on the same network, without needing to have a VPN set up. I go, yep, sure enough, that VM looks like it's consuming more storage than I would have expected. So I'm going to go right into the VM console here. I will copy and paste my password into the console, because I don't want to have to type that in. And then from there, we'll go into the VM. And I'll look around. Oh, how fortunate. I have a folder called unnecessary data. So I'll look at that. Sure enough, I can delete some of these things. I'll go ahead and delete those items to reclaim the space, because, again, it's unnecessary. I want to make sure my clusters are running as efficiently as possible. So I'll just delete those in Guest. And then, of course, HyperCore is intelligent enough to accept that unmap or discard command issued automatically from this Guest operating system. And so just within a few moments, I reclaimed a lot of that space on this VM. And now I'm looking good and can go back to my daily tasks. Awesome. So Mitch covered several things here. So discard support is console paste and VM metrics. So all this stuff is really soon. So we'll go through quick details here. Card is in 9.6. Two big things here. It frees up the unused blocks inside the OS to make that storage available for other VMs on the system. And it also aligns the utilization metrics. So what you see on the disk and card in HyperCore and the VM metrics in Fleet Manager, that's going to be much closer to what's actually happening inside the virtual machine. We know this is a feature people have wanted for a long time, getting passwords and software keys and encryption keys into the console has been very painful. This is coming in 9.6. And of course not only does this work in HyperCore, but it works in Fleet Manager through SecureLink. And as of this week, VM metrics are live in Fleet Manager. So this lets you really easily dig in to see how VMs are trending over time and which VMs are using more resources than they should. This is a great troubleshooting and resource planning tool. So Taylor, you're actively managing 1,000 locations. Can you tell me a little bit about how scale is helping you with your day-to-day? Yeah. Absolutely. So with 1,000 sites, we also start our day in Fleet Manager. We're using dashboard to aggregate the views for all those sites. But with a limited staff and, again, with that need for 24-7 uptime across the country, we have to also rely on automation. So we took pretty heavy advantage of the new Fleet Manager API. We're doing a couple things with it, but two of the cool things I wanted to call out are we have a scheduled job that retrieves all the conditions. So instead of retrieving them from 1,000 clusters, which you could have done, now I just hit the one Fleet Manager API endpoint, and that can auto-create and auto-close cases in our case management service, which happens to be ServiceNow. The second thing we did is we used a little AI assistance to do some utilization monitoring. So we pull those VM metrics that Mitch showed in the UI. You can also retrieve them from the API, and you can run some basic analysis on it, and you can actually predict when CPU or, of course, storage is projected to run out. And we have those alerts sent directly to an IT Slack channel. And real life, Taylor, you did this, right? You just got into AI, hooked up, got the APIs, and you wrote this script. Yes. And you're not a coder, right? No, I've coded it. So anyone in the room can do the same. Nice. So the Fleet Manager API is available today. So if you go into Fleet Manager, you can get access to this right now. And again, there's tons of APIs here, lots of cool stuff. And Taylor's going to talk more about it in more detail tomorrow, or sorry, later today at 2.35. All right. All right. Now, Taylor, I'm the CEO of this pharmacy. I know you like your little charts, but I need you to deploy AI right now. Wait. Why? Why now? Haven't you seen the news, man? It's like all rainbows and unicorns. Just do it. Just do it. All right. So Mitch and Taylor have both gotten an edict from their CEO to deploy AI. Mitch, you need to deploy a new app. Can you walk me through that? Yeah. So I can't believe we have the same CEO, by the way. That is unreal. So we want to deploy an AI-enabled application that improves our quality assurance on the manufacturing lines, and it does require GPU resources. The original hardware that we specced out didn't include GPUs, so the first thing I need to do before deploying this app is add additional resources to my cluster. So here I'll go into HyperCore through the control center to the control section, and you'll notice there is an add node functionality here. So first I'll go ahead and click add node. From there, just specify that node's configuration details, click add yet again, and HyperCore then will automatically, non-disruptively add that node to the cluster. So no interruption to my existing VMs, and, you know, after a few moments, that node will get added, and from there I have a fourth node now with GPU resources where I can deploy this new AI application, and, you know, we're really excited about the business benefits of deploying this app, increasing our quality, and then seeing the downstream revenue benefits. Nice. So Mitch talked about the fleet manager and HyperCore, node lifecycle management, node out will be part of the UI and part of the full node lifecycle management, so you'll be able to add nodes, delete nodes, replace nodes, all right from the UI, and, you know, I think as we were talking about, or he was talking about before with the friction, right, now that's a call to support that's no longer in your way, right? You can just do it yourself. Until recently, GPUs were really a niche use case for, you know, sort of extreme VDI situations, but with AI, obviously, we're seeing demand for GPU everywhere. We do have a GPU-enabled platform now, but it does have some limitations, and we're committed to making GPU a first-class fully supported feature in HyperCore, and we've invested in a team to make that happen, and with that, live migration support on GPU will be coming in Q3, and then onboard GPU and mixing and matching GPU nodes with non-GPU nodes will follow. We have a bunch of cool stuff on the hardware side coming. We have a refresh of the HE150. We have the HE250 from SimplyNUC, which we've been shipping for a bit, and we're happy to announce that HyperCore is the first hyperconverged infrastructure stack certified on the SE100 from Lenovo. You can fit three SE100s in a 1U rack. It's super, super cool. It's out at the Innovation Center, so I would definitely stop by and check that out. Actually, all the hardware I talked about is out there, so go buy, touch it, feel it. It's super cool. I'm really excited about that one. So one of the most exciting things, I think, happening at scale right now is Project Kraken. Jack talked about this earlier. It enables application delivery and management across your fleet. You can deploy applications to multiple clusters all at once with a click, and the applications get packaged as a manifest, so it gives you infrastructure as code, and that includes all the assets, configurations, scripts you need to deploy not only the VMs or the containers but the applications inside of that, and this gives you strong consistency across your whole fleet versus a manual approach. So Taylor, you've got to get this AI application deployed across 1,000 sites. We're talking about Kraken, so I'm guessing you use Kraken, but why not Terraform? Yeah, so obviously we'd already shipped them out. They were successfully installed across the country. We don't have VPN access, so after the sudden directive from above, I was very relieved that we'd selected that AT153P platform that has those onboard GPUs that we could leverage, and so I was just going to show a quick demo of how we deployed this Computer Vision app across those doors. So in Fleet Manager now, there's a deployments tab, and I can select any number of clusters. I can select any number of applications. In this case, a Computer Vision application. It's actually a Docker container in this manifest, kind of like a Docker file, and I go into Fleet Manager, and I just hit deploy, and using remote access technology, those clusters which currently don't have these VMs, and in fact, don't even have the media required, are now pulling down the media. They are booting themselves up, and now they are appearing in Fleet Manager. And I can go and I can grab the IP, and just like that, I now have Computer Vision deployed in the faraway pharmacy. Yeah, this is awesome. So this is why we're really excited about Kraken. We're taking this application deployment thing to the next level, and like Jeff was talking about, this lowering the bar for deploying an application out to the edge is lowering the bar for innovation at the edge, and that's really what we're trying to do for everyone. So Dave Dimlow and Jack Hall are going to talk more about this tomorrow afternoon. They're also going to be talking about how we work with containers and all kinds of stuff, so Kraken, containers, it's going to be a great talk. Be sure to check that out. So I talked earlier about the getting early access to 9.6. So we're starting a new program called the Early Access Program, and you'll be able to get early access to versions of HyperCore before they become full production, or before they become full release to everybody. If you have a test lab or a test cluster, a lab cluster, a pilot cluster, anything where you're comfortable deploying something a little bit early, it will be full production support. You can call support, get help, it's not like a beta. But if you want to test early versions of HyperCore, this is the way to do it. There's a QR code here to walk you through it, but it's super, super simple. You just go into Fleet Manager and add the HyperCore-early tag to the cluster, and once there's a new release available, you will see that as an available update for that cluster. And the tag for that cluster, again, HyperCore-early, and there's the link again. So don't forget to check out all the things we talked about, a bunch of cool stuff happening in the next couple of days. And tomorrow afternoon, Jeff threw this up, Jeff and I are going to do a talk about where things are going over the next five years. If you have questions, grab this link, type them in, and we'll get to a few of those for sure. So one of the best parts of my job is to get up here and talk to everybody about all the cool stuff that's been happening over the last year, but I wouldn't be able to do this without my team. So would you please join me in giving Scale's product team a big round of applause for all the hard work they put in on FCProperty. Thank you.
