Infrastructure Matters on the road at Cloud Field Day – Infrastructure Matters, Episode 29

Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this webcast. The author does not hold any equity positions with any company mentioned in this webcast.

Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.

Transcript:

Steven Dickens: Hello and welcome to another episode of Infrastructure Matters. You’ll notice a different background and some different people on the show today.

Camberley Bates: Interlopers.

Steven Dickens: Interlopers, yes, we’ll call them interlopers. So, let’s do some introductions. First off, to my left, we have Stephen Foskett.

Stephen Foskett: Thank you for having me here on Infrastructure Matters. I would not dispute that statement. And I am the organizer of Tech Field Day here, now part of The Futurum Group

Steven Dickens: And then the other guest on our show.

Keith Townsend: I’m Keith Townsend, formerly this President of the CTO Advisor. Now I’m just a head advisor here at Futurum Group.

Steven Dickens: Hanging out, chilling.

Keith Townsend: Hanging out, chilling, enjoying. It’s weird. I’m at a Tech Field Day. We’ll get into what that is, but I’m at a Tech Field Day. I’m not a delegate. It is really strange.

Steven Dickens: So let’s go there first. We’re on the road at Tech Field Day. Fantastic first day of that. But maybe for some of Camberley’s and I’s core listeners, they don’t know what Tech Field Day is. So Steve, tell us what it is?

Stephen Foskett: Sure. So Tech Field Day is something that I started 15 years ago where I-

Camberley Bates: Fifteen years?

Stephen Foskett: Fifteen. Can you believe it? No kidding. No kidding. I know that because-

Camberley Bates: I’ve known you that long? Oh, my God.

Stephen Foskett: You’ve known me way longer than that.

Keith Townsend: I know it’s 15 years because five years ago there was a 10 -year anniversary.

Stephen Foskett: Yeah. And that was a big party.

Keith Townsend: It was a big party, great party.

Stephen Foskett: So basically, a bunch of us who were speaking at events and getting champion, MVP, that kind of thing, and we were starting to get invited to things, as blogger days. And because we were writing blogs, and we’re on Twitter, and all that stuff. And so, we decided to organize our own. And so, essentially, instead of having one company invite a bunch of people in, we invited the companies to present to us. And so, over a couple of days we would have six or seven companies come in. They would present to us, we would ask questions, we would attend, as if it was a blogger day. And one of the things that we decided to do was to start live-streaming, and recording it, and sharing those recordings.

That became a real valuable thing in the industry, because essentially now, on YouTube, there’s literally thousands and thousands of these videos of people like myself and Keith listening to presentations, watching demos, asking questions, engaging. Now that we’re part of Futurum, it’s pretty cool because essentially, we get to have people like me and Keith, but also people like the two of you, around the table asking questions, engaging, interacting, sharing your thoughts. And it is all wide open for anybody to watch.

Steven Dickens: So you’ve been a long-time delegate? That’s not a phrase that I think makes sense straight away. Do you want to explain what being a delegate means, Keith? Because, I wasn’t aware until I came today, I was on the pre-brief calls, I was on the calls, I heard the phrase. I’m like, “Okay.” But maybe just explain what being a delegate means and what that role here is to represent the end user community.

Keith Townsend: So, I love it. Stephen tries to have at least one new delegate every event. It’s tough. It’s not always possible. But we had at least one new delegate for this event, well, three new delegates.

Stephen Foskett: Actually, if you count these two, we have five…

Steven Dickens: Wow.

Stephen Foskett: … which is absolutely incredible. It’s been very rare that we’ve had that many new delegates. But it was funny, the first one, it was all new delegates.

Keith Townsend: That is how that works. Yes, yes. So, my first Tech Field Day event was way back when Storage Field Day Five, which was in 2014, I think?

Stephen Foskett: Something like that.

Keith Townsend: Something like 2013, 2014. And I’ve now done 35 of these…

Steven Dickens: Wow.

Keith Townsend: … over the-

Camberley Bates: So okay, but what is a delegate?

Keith Townsend: So I was just about to say, I think the best way to describe a delegate is, a independent influencer, whether they have a voice via their blog, website, social media, really doesn’t matter. But they are somehow participating in the community. They can be an analyst, they can be an end user who’s involved in the community groups, or like me, at-large at the time, at-large influencer who made money full-time, just talking tech. Extremely technical, like the two of you. Able to digest this at a much lower level than a typical analyst. So think of an Analyst Day and you’re an analyst, but you’re talking deep technical concepts. You are the voice of the community of practitioners.

Stephen Foskett: And I think that’s the key point, there. Ever since the very first one, I realized that the role of what we now would call influencers is not in that direction, necessarily. It’s actually in the other direction. The delegates, the reason we call them delegates, right from the very first event, is because I realized that the people around that table, they represent the people who aren’t around that table. They represent the people who want to get their questions answered and want to learn this stuff. And to me, that’s the core of what we’re talking about. The voice.

Steven Dickens: Yeah, that’s come home from me today with the representatives of a broad, with the kind of, I’m trying to embody it today of, “I’m asking questions from a group of people who maybe think like me, got skills like me, experience like me.” My job on one of the, what are we 12 today? Is to ask those questions because I’m in the room. But I’m trying to think, I’m asking those questions from a bunch of people who are where my head is that trying to understand. So, I’m in the room, and I’ve heard you ask a bunch of questions the same today, Camberley, from that Storage mindset, you represent that community really well.

Camberley Bates: We’ve been doing this for quite some time and representing the IT and user community. So, I think that it was a natural flow for me in some of the questions that we had because how our analyst group had operated before we were part of Futurum. So, I might object to your statement about the technical stuff. Although, I’m not quite the practitioner you were, that’s for sure. But there’s an element, and I think that we’re bringing into the group that’s both like, “How are you going to get to market? What’s this pricing like? Is this going to get adopted where this stuff…”

And also that piece of it, the technical piece of it, the practicality of it, “Is this really going to work for me,” that I heard. And then, I also heard some really deep questions that we’re like, “Okay, so that’s an interesting one to where that’s going in terms of the integration piece of it.” You could literally hear the pain point when they’re talking, what they had bumped up against and where this hits on the pain point. So, yeah.

Keith Townsend: That’s a good point. I really love the interplay between how both you and Steven take a very industry-wide look at the presentations. You talk to a lot of vendors. And you’ve been briefed on the high level. And you came with some really great questions around where these solutions fit in the markets. And then, hearing the practitioners come back. And a couple of the practitioners were really scratching their head, like, “I don’t understand the problem that this solves.” And as analysts, it’s really easy to get wrapped up in the high-level talk track. And not have that really low-level context that some of the other delegates had of saying, “You know what, this is a people problem being solved by a tool.” And we heard a lot of that today.

Steven Dickens: Well, that gives to me a perfect segue. Thank you Keith, as always. Great assist. Let’s dive into some of the vendors that we’ve heard today. We’ve got three vendors on the docket. Maybe, let’s start with, you know the agenda better than me. Who was up first? I’m a bit-

Camberley Bates: NeuroBlade…

Steven Dickens: NeuroBlade.

Camberley Bates: Was up first.

Steven Dickens: Yes, the wonderful NeuroBlade. So, I think coming up, this is Cloud Field Day 19, a particular set of requirements. I’d be keen to get your view, maybe I’ll add some color, Camberley, if you want to go first?

Camberley Bates: Yeah. And this gets back to what a little bit of what you were just saying, is that we’ve seen the XPUs, DPUs, things coming up over. Now, that group may not have seen as much of that come to market. We’ve also seen the DPUs come to market and just struggle to get to market, for the same things that we are raising with them, which is, “What’s the level of integration? What is it going to take to get out there? Where’s the route to…” those kind of things that have struggled with those folks getting out there.

But then also, seeing the light bulbs, and the team, and the delegates going off and saying, “Okay, so this is what this is doing, 30-times faster,” that’s unbelievable, which we have problems with those kinds of numbers, as well. Prove it to me. Which is the same thing that we talk about. We’ve talked about. So, I think it was a real, well-rounded discussion. I think NeuroBlade, from my perception, has got something there. And they’ve addressed some of the big issues that are the barriers to market, in terms of the integration problem, and the systems, and various focused.

Steven Dickens: I’ll just, for the listeners who didn’t see-

Camberley Bates: I didn’t even say what it was. I’m sorry. I didn’t even say what it was.

Steven Dickens: So it’s a SQL processing unit. These guys have developed custom silicon to insert into that SQL processing thread or architecture, to be able to accelerate that. So if you think of a graphics accelerator in your laptop accelerates the graphics processing. This is the same model, but comes in to accelerate those SQL queries. So custom silicon working with a fab partner to create a card, literally yay big, that goes into a server. What were your key takeaways, Keith?

Keith Townsend: So I couldn’t help, Steve is going to chuckle a little bit about this. I was listening to it on the way walking over. I didn’t sit in to the whole presentation, so I didn’t catch the whole thing. But I got the gist of it and it threw me back in mindset to this presenter back, it had to make Storage Field Day Eight or something like that. And, they also had a SQL accelerator, it was a driver that slipped into Windows. And they called it the, well, we labeled it the Blackest of Boxes. We didn’t understand how the technology worked. And I really appreciated the time that NeuroBlade took to explain how the technology worked. I think, Camberley, that caught one of your questions on, is this basically pre-loading the stuff into memory and solving the RAM problem of an individual system. And you instantly get this. “Okay, pseudo in-memory database.”

I’m creating an in-memory database using an accelerator card. So, well, maybe not 30-times faster. Again, we’d have to validate that. But, the ideal of it being significantly faster and worth the overhead, the operations overhead of deploying these things, and making sure they’re available. I didn’t catch and see, “Well how would this work in my VMware environment? Is this invisible to VMware, et cetera.” And those are the questions. So I’ll go back and watch the sessions, because I’m sure the delegates asked these questions or they addressed.

Steven Dickens: Oh, yeah.

Camberley Bates: They did address the VMware and the container space. So, yeah.

Steven Dickens: What were your takeaways, Steve?

Stephen Foskett: Well, I was really impressed by them. I met them at FMS last year. And I was really impressed by what they’re doing because it is so practical. Like Camberley was saying, the reason that I think some of these accelerator cards sometimes don’t take off in the market, is because they’re a solution in search of a problem, number one. And number two, because they’re hard to actually make use of. And that’s-

Steven Dickens: There’s barriers to adoption, basically?

Camberley Bates: Big time.

Stephen Foskett: And of course, those of us who’ve been in computing for a long time have seen this again and again. You get all these accelerators that are deployed. And if there’s not an easy way to use them, then they’re just not going to be used.

Steven Dickens: Yep.

Stephen Foskett: And so, the thing that I like about NeuroBlade is that it really is transparent. It fits right into the SQL workflow. And again, a database application, it’s going to be a higher-level application than storage, or something like that, or even graphics, or AI. And it makes it a little easier to slot it in there. They’ve got good software support. And what they’re doing isn’t really magic. It’s the same parallelization and pipelining that we do in A GPU, except that this has a very specific use case. And that specific use case is SQL data applications that are just too big to fit in memory, too big to ever fit in memory. It makes a lot of sense to me. I really like it.

Camberley Bates: And I think the other piece is that they’re very focused on the data analytics space. So right now, they’re with Presto, which is a key application, and working on Spark. So they are in the core of some of those analytical analysis places going. And I’m actually wondering, now that I’m thinking about it, since Spark is so prominent in the security industry, and tracking that, I’m wondering if that’s really a place that it can address the market, in terms of the analysis of the database? We look at, that’s all that data coming in there for ransomware and what they’re getting attacked to. That might be an incredible app for them.

Stephen Foskett: And I have a suspicion that that’s the applications they’re already being used in, because they couldn’t really talk about their customers.

Camberley Bates: We can’t talk about them.

Stephen Foskett: And I don’t claim them.

Camberley Bates: “They’re the three-letter letter people,” or something like that, is what they were doing.

Stephen Foskett: If you imagine the kind of people who need to accelerate Presto, and Spark, and-

Camberley Bates: The banks.

Stephen Foskett: Yeah, it’s financial, it’s intelligence.

Camberley Bates: Yeah. Yeah.

Stephen Foskett: Okay.

Steven Dickens: People are going to build.

Keith Townsend: They mentioned healthcare, I’m like, that seems like a throw away. I’m like financial security.

Camberley Bates: They kept staying away from that as far as they could. And it’s like, “Okay, let’s go down this one.”

Steven Dickens: I think that’s the financial services question and they managed to dodge it.

Camberley Bates: Three times. It’s like, “let’s touch it.”

Steven Dickens: So, next on the docket was SoftIron. Really fast, I’ve been tracking those guys, I spoke to Phil Straw, their CEO, on one of our podcast channels 18 months ago. They’ve briefed me probably three or four times. Really interesting to watch a bunch of practitioners who are deeply invested in the software stack, VMware, Kubernetes, platform engineering, how you make this stuff work, listen, and react to a platform that’s completely rethought that from the ground up, from the silicon to the user interface, and not taking off-the-shelf pieces, and parts, and products, and meshing them together. So it’s fascinating to, because I was one step ahead of you guys, because I’d seen it. And it was fascinating to watch you guys go through the same reaction like, “How does that do…,” So maybe just react if you will, to what you saw from SoftIron?

Stephen Foskett: Who do you want to start with?

Steven Dickens: We’ll go with you, because you’re next to me.

Stephen Foskett: Okay. I’m next to you. Okay, I’m glad. So, I think the thing that there was a lot of light bulb moments with SoftIron, with the crew, like you mentioned. Number one, that this isn’t just another hyper-converged infrastructure stack, woohoo, right?

Steven Dickens: Because that’s what it looks at the first time you see it, like, “I’ve seen one of these before, it’s hyper-converged. Oh, 20 people have done that.”

Stephen Foskett: Yeah.

Steven Dickens: And then they get to charts two, three, four, and it looks completely different.

Stephen Foskett: Yeah, it’s a compelling, it scales way further than any hyper-converged stack does.

Steven Dickens: The answer to your question about scalability was the best reaction of the day. We’ll come back to that, Camberley.

Stephen Foskett: And so, that was one thing. Another thing was when the light bulb went off about why this is hardware plus software? And they’re like, “Oh, so this is hardware for people who can’t buy off-the-shelf hardware, due to security concerns and things like that.” That was a big light bulb moment. I think another big light bulb moment was when people realized that this thing was originally designed in the ARM space, that they’ve been developing this thing, and they’ve been enhancing it. And it wasn’t, again, it’s not just, “Let’s get a bunch of Dell servers, and install some software on it, and virtualize some stuff.”

Camberley Bates: Well, okay. So I’ll take a piece of that. You do have the three-letter companies. You do have the Feds, you do have the DOD, all those guys are buying Dell and HP servers.

Stephen Foskett: Mm-hmm.

Camberley Bates: And they’re also working on some big Zero Trust capabilities very closely with those organizations. The difference here is that, and where I think the difference is, they have taken it to another level of building. And that other level says, two areas. One, they’re looking at each and every component that’s in, whether or not it’s a network node, or a storage node, or server node, and say, “What do I need specifically for my environment? This environment that we’re building?”

Steven Dickens: It’s a known fact.

Camberley Bates: Not a general purpose box, or even if I’m done, focused on, every one of the server guys will build a box that’s specifically for analytics. So they’re specifically for VM stuff. But there’s still, there’s a level of generality to them, because they’re hitting a mass market. These guys are hitting a very specific market. They’ve built from the ground up everything but the chips. And I presume that the solid state drives or something like that a Zero Trust Environment, it’s not done in Taiwan, it’s not done in some of the other places that everybody’s concerned about doing. So you got that piece. And then they built their own software stack. Okay? And now you’ve got that entire thing integrated. There’s a whole lot of questions that I still have. I could spend a day with them going through this architecture, like a good architectural white paper to talk about, “How does this thing scale?,” because they’re talking about into the petabyte range in a single cluster.

Steven Dickens: Your Reaction to when you asked the question about scalability, they came back with…

Keith Townsend: I think they said something like 16 petabytes.

Steven Dickens: … 16 petabytes. And your reaction was like…

Camberley Bates: Well, it’s double-digit.

Steven Dickens: … 400 questions just by association.

Camberley Bates: Right. Everybody in the group did, because we talked about, the issue that we talked about is that when you have hyper-converged systems, you’ve got blast radius problems and you have scale problems. And, for the most part, most of those systems are going up to maybe eight, maybe 16, in that environment. And at that point, despite the fact that we’ve got, maybe separate nodes that can scale out, and, yeah, the Nutanix that scale out, and those guys will talk about that. There’s still scale issues, with the exception of probably one of the pieces of software that’s out there. And that’s power scale.

So this is really an interesting, something I really want to get into farther, in understanding what they’re doing and how they’ve been able to pull this off. And there’s this unique piece that they have, that’s part of the network, that’s part of the network nodes, that is doing the control plane, it’s like, “Okay, the network nodes are doing the control plane for the entire environment.” That one took me for a loop.

Keith Townsend: I had a ton of questions.

Camberley Bates: And about two of them got answered.

Keith Townsend: About two of them-

Camberley Bates: Because we didn’t have time. It’s not that they, we didn’t have enough time, and there was a whole lot of juice.

Steven Dickens: We all get 90 minutes for three of-

Camberley Bates: It’s a lot more juice in this, Steve.

Keith Townsend: I would love for our Labs Team to get this in-house, and tear it apart, and understand it. I think, the one thing I was surprised at, was delegate response. We’ve seen these cloud-in-a-box solutions for years. And they have not taken off. They just haven’t. And it’s not about the lack of features, it is about logistics, and buying culture, and operations. One of the delegates made a really great point about cold termination of VMware licenses with hardware that typically doesn’t happen. Not only does it not happen, this is a different operating model for operating infrastructure within a enterprise IT shop.

So the question is, I made the comment in Slack that this is the thing that customers think they want. Then when you show it to them, they get excited. And then when they go through the details of what it takes to implement it, “Hmm. Wait, so how do I do backup and recovery? How do I do replication? How do I do storage? Are my apps designed to take advantage of this?”
So, you talk about standing up a new pod. What I was surprised about with the delegates, in addition to this, is that they didn’t, of all the solutions we’ve seen in the past, they like this. The response was, Michael Levan?

Steven Dickens: Yeah.

Keith Townsend: Levan is really tough when it comes to Kubernetes space. You can’t get much over it. You can’t get much over this.

Stephen Foskett: He’s on the Kubernetes Dev Team.

Keith Townsend: Yes, he is intense. So, if you’re going to come with anything related to Kubernetes and you’re presenting to him, make sure you know where the value was at. And I was shocked to see that he really didn’t poke many holes into the solution. So technically, it sounds like a sound solution. I’m just wondering what is it I’m not seeing? Because this is a huge, massive undertaking,

Steven Dickens: It’s huge.

Keith Townsend: To do this whole software stack in the hardware stack.

Steven Dickens: For me, and they touched on it at the start of the presentation, they’ve gone and got funding in a different way, because they knew this wasn’t a five-to-seven year journey. These guys are 12 years in. And are just getting product market fit. So the thing that’s been fascinating to me is, these guys knew it was really hard to completely re-engineer this from the get-go. They knew that it was a complete rethinking of the hardware, the silicon, the BIOS, right through to that control plane and everything in between. So they’ve taken the long-term, “How do we rearchitect it?”

And that, to me, is really thoughtful. And I think, as we now start to put it in front of people, the questions come. And because they’ve taken 12 years to get here, they’ve had time to make missteps, they’ve had time to have it be picked apart. And the thing I’ve picked up from SoftIron is just the completeness of vision. So, keeping us on track here and keeping us moving, it was really interesting to see our next vendor come up off the back of platform…

Stephen Foskett: : SoftIron.

Steven Dickens: … SoftIron and then Platform9, because they won’t come in at exactly the same problem. But there was elements of that. So maybe, Keith, you know this space maybe better than anybody around the table. How did you see Platform9? And how would you describe it to the viewers?

Keith Townsend: So, Platform9 has been around for just over 10 years. I believe. The team, the founding team came out of VMware. There was a product called VMware Labs, which I used and was beloved. The team started Platform9 as a managed OpenStack solution, where you would basically give them, install their agent on VMs, bare metal, and then they stand up an OpenStack instance within your environment, and they manage it. That has now morphed, of course, to Kubernetes.

Steven Dickens: Yep.

Keith Townsend: And they’re trying to solve their latest product announcement. They’re trying to solve a problem that has existed in the enterprise, since, I’ll just say since the time that x86 servers entered the data center. The over-provisioning of resources and the inefficient use of resources, they threw up a slide that was not a surprise, I think, to any of us, because we see that number quite commonly, 30% of the resources in a given data center are used. So efficiency of a data center is roughly about 30%. And that’s a pretty common number.

Steven Dickens: Yeah.

Keith Townsend: What they’re aiming to do is increase that utilization using software to, basically, re-provision EC2 instances as needed and right-size those instances based on the need of the application. There was a lot of debate within the delegate class back to Platform9. Basically, are they trying to solve a people problem with a tool?

Camberley Bates: Yeah.

Steven Dickens: Yeah. The one that stood out…

Camberley Bates: That was the big-

Steven Dickens: … for me was, you got your application needs 10 gigabytes of space, but minimum you can provision in the instance is 12. That doesn’t sound like a very big problem when it’s two gigabytes on one instance. And I made the comment, I think, you times that by a thousand, you’ve now over-provisioned a thousand environments by 20% and you’re paying for that. But going back to your point is, is that a people problem we’re trying to fix, or is that a technology problem?

Camberley Bates: We’ve been trying to fix that problem for a very long time. And the balance of the developers, or the database guys, or the storage guys, or whatever has always been there. So we’ve applied new types of technologies to that. On my storage side it would be thin provisioning. We’ve done virtualization. We’ve been doing this since MVS. So, it’s not new. It’s applying it. If you’ve got a super-smart platform engineer, you got a super-smart delegate, like we had around the table, yeah, you’re going to address it. But we don’t have those kinds of skills.

So now, I’ve got to look at it and look at it another way. So now what they’ve done is, outside of what they do from a managed servicing, they brought a tool that you can bring to AWS, I think they plan to bring it to shore, that you can do saving of the servers and get your money back. Some significant dollars on your money. So, good stuff.

Steven Dickens: What was your takeaway from those guys, Steve?

Stephen Foskett: Well, I’ll just say this, that you’re right, of course, that this is what we’ve been trying to solve in various areas.

Camberley Bates: Yeah.

Stephen Foskett: Thin provisioning was actually a pretty successful…

Camberley Bates: Yeah.

Stephen Foskett: … product category.

Camberley Bates: Yeah, very category.

Stephen Foskett: But not nearly as successful as VMware. Which…

Camberley Bates: Right.

Stephen Foskett: … the main, honestly, the main slam dunk use case for VMware was this problem

Camberley Bates: Was server utilization.

Stephen Foskett: Server utilization. And they’re doing it…

Camberley Bates: We’re back there again.

Stephen Foskett: … for the cloud. And so, I guess you could do worse than attacking the same problem in a different domain, with the same solution that led you to be a multi-billion dollar company. So, I’m liking it.

Camberley Bates: Yeah.

Steven Dickens: That was my takeaway.

Camberley Bates: Yeah. But, yeah.

Steven Dickens: There’s some challenges, I think, around their shared-risk/shared-reward model from a pricing point of view.

Camberley Bates: Oh, they’ll figure that one out.

Steven Dickens: I think they were trying to maybe, whilst I applaud what they’re doing to try and share the risk, that’s going to be hard to sell to people in procurement. But I think you can knock the edges off that. The core engineering looked solid and that’s a lot harder problem than coming up with the pricing model.

Camberley Bates: All in all, good session.

Steven Dickens: So yeah, we’ve had a fantastic day here at Tech Field Day, Cloud Field Day 19, I think we’ve had. So check out the content. You can see that streaming on the LinkedIn Channels. You can check all of us out on our socials, we’ll be posting about those sessions. So, we’ll see you next time for another episode of Infrastructure Matters. Thank you very much for watching.

Other Insights from The Futurum Group:

Platform Engineering and FinOps Converge in Kubernetes Deployments

The Evolving Landscape of Private Cloud Infrastructure – The Futurum Group

NeuroBlade Announces New Partnership with Dell Technologies to Accelerate Data Analytics