Roundup on Tech Field Day for AI Infrastructure, Nutanix .NEXT, OpenAI, RSAC – Infrastructure Matters EP81

All right. You are watching, listening, or however you're consuming. You could just be reading the transcript of Infrastructure Matters.

Episode 81. I'm here with my good friend Kimberly Bass, who's head of career transition, but is still beating me to joining the stream every time. Kimberly, welcome back to the show.

Uh, Thank you very much. Yeah. Um, for those of you who don't know, I've, uh, officially have gone to the, the next side of my other third, the next third of my life.

Um, and I'm, I'm not working full-time with, uh, I'm not working with Fu and Group, but I'm still having the opportunity to participate in this Infrastructure Matters, um, podcast, which is just a great, great fun time, um, with you guys. And I'm missing Dion. He's, he's, he's someplace we are not sure where he is today.

If you dunno where in the world Dion is, I know we're about to get into travel season, I'm sure. SAP Sapphire is coming up quickly. And Dion is a staple for SAP Sapphire.

Wow. Uh, which, uh, usually coincides with Click Connect, which, which I'll be at the week that you're watching this program. Uh, I'll be at click Connect.

I be believe Dion will be at, uh, SAP Sapphire. And then, uh, following that is Dale Tech World. So we're getting into the swing of the enterprise.

It kind of spring, early summer schedule. I'm expecting to receive a new toy from Dale today. Actually.

They, they, Dale doesn't like what, when us Analyst type show up at their events without a, with a MacBook or hp, uh, laptops, or they're sending us all laptops to make sure that we're uniform. And the Dale. Mm.

They never sent me one, but, you know, I've got my, they've Never sent you a laptop. This is the second laptop I gotten from Dale in the past six years. So, uh, and, uh, I'm not gonna tell you what I do with the laptop.

I don't sell it 'cause it's technically not my laptop. It, it is a loaner, so I can't sell it. But I absolutely keep it and put it to good, good use.

Well, what, what we do, uh, maybe they could send me a PowerMax or something. We'll see. You know what that, that will make up for the years of them not sending you Yeah.

A laptop. Definitely. So, So let's jump in.

Let's do, uh, let's do a quick recap 'cause we missed last week. The team here, uh, at Techron that typically produces these videos was off at RSA. And we'll get into some of the kind of talk of RSA because it is a big security conference.

You though, either even though me, you nor Diane really cover security. Maybe we need to get Krista on Yeah. To talk security.

But, uh, the week before that you were at Tech Field Day, AI infrastructure day, or AI infrastructure, uh, field day. Any big themes you wanted to share with us coming out of that event? Well, so this is the second AI infrastructure field day we've had.

Um, and the first one, we, we more focused on the data side of the house. And this one, we opened the aperture to, um, various technologies that came in the door. Um, and that included everything, a lot of networking.

So there's a lot of fo focus. Um, Juniper was there, um, even, well, we spent a full day at Google's offices, and there's a great video that was put up, um, that Colleen call had done on me. It was really sweet of her.

Um, but talking about infrastructure matters, why infrastructure matters, and I, you know, it came away from the Google session, which was an all day session, including the networking and the data and the servers and the storage, and a lot on the storage environment saying that they're not trying to put you in one, one box fits all, um, or one cloud fits all or one cloud kind of environment. They understand that this is a very complex environment, that you're doing different things with data. You're, you know, you're doing l lms, you're training, you're fine tuning, et cetera.

And so that there is a, a myriad of solutions that they're bringing to you. Um, and it's not for the faint of heart. So I think that having a consultant on board to help you through how you're designing what you're designing is really stupid important.

Flip that to the next p. One of the other ones that was really interesting, which was Nutanix that came in with what I would call the easy button. And, um, Nutanix, they're known for easy button, right?

You know, HCI stuff, um, hyperconverged. And, um, and what they've done is they've put, they've gone and curated like the HuggingFace, um, models and selected the top of the models. There's, you know, thousands and thousands and thousands of models.

So you can get buried in there to say, what's the best model for you to be using? And with that curation, you can select a model within the, um, Nutanix box, and then it automatically populates your GPU, your VM environment, all those kind of things for you to kinda get going and get running. So, you know, there's a bunch of easy buttons they've done with it.

It's pretty slick. Um, and, um, I think that it's designed for the companies that say, okay, so I wanna just get going and try this out, get this working. Um, and yeah, suppo, you know, it scales.

I don't know what the performance is. Usually you give a little bit of performance out for those pieces of it. So that was really interesting as well.

So those are a couple of the highlights. Um, we heard from our friends at Soy, and, uh, they came in and you can talk again about their, um, their, uh, liquid cooled technology, which had everybody like odd, we were, we were all like in the, the petting zoo. They were passing around all the, the, the fun toys to us.

And it's like, Ooh, ooh, ooh. Yeah, it, it, you know what the delegates really like, uh, when soy presents the back channel, I'll, I'll tell a little secret from the field day back channel, when the delegates get together and talk, they said, how is it that a storage company is coming in and, and explaining AI better than some dedicated AI companies, solid Im does a really great job. And it's not just a storage company, they're actually the device.

I mean, they're making the solid state drive device. It's not like, it's not like they're out there, you know, like a WCA or somebody like that. I mean, they, they outdo the, the guys that put the storage systems together, um, in terms of their understanding of this entire, the complexity of what it is.

And so, and they are building, and the reason why they've done that and gone deep and really good at what they do is they've gone deep to understand what are you gonna need from an IO capability, capacity, capability at every step of the way of delivery. And, um, they've done an excellent, excellent job of that. Yeah, the, it reminds me, I have this saying when I'm mentoring, uh, career professionals, especially independent contributors, you need to understand your value to your organization all the way through the product that the product or services that you deliver.

So if you're a DNS administrator, you need to be able to explain to the, uh, CFO why your job exists beyond, you know, doing, making sure DNS runs. Mm-hmm. Same thing with, uh, basically what Soy did at the presentation, if you haven't watched it, they absolutely understand their value chain.

And it's a great just lesson. And not just marketing, but product, you know, product engineering. Like what, what role do I play ultimately in making AI real?

And they did a really great job. So head tip to the folks at the solid, And one of the other companies that was there, which is also a process company or a, a, um, component company is Fon. Um, Fon hasn't been really well known.

They're, they're almost $2 billion company. Um, but they are the, um, the drives solid state drives behind, uh, a lot of other people that, so they have put other people's names on it for specifically, if you bought solid state through Seagate, it would've been coming from fisa. Now recently, they've gone pub, they've gone public with their name.

Um, we hear a whole lot more about 'em. And then what they were there to really talk about was their adaptive, um, software technology that, uh, goes with the drives and enables you to more efficiently use, um, HBM or the, the memory and capability when you're doing the training. Really, really kind of slick stuff.

There's a blog up there that I'll, I'll put that kind of summarizes what it does, um, and the capabilities that it brings to va. So it lessens the amount of memory that you have to have, um, in order to train, which is really significant because that's one that's probably the biggest, um, bottleneck right now that we have with training among all the other pieces of it. Um, so those are a couple of, a few, a few of the folks that, you know, I would like to, you know, highlight in here.

That was, uh, really a good session. Um, so I'll turn it back over to you, Keith. Yeah.

And, uh, you mentioned these drive riders, you know, kind of the, um, the big claim of fame was kind of the 122 terabyte drive, right? And one of the things that we missed because we weren't on last week, was Dell announcing their latest, uh, AI server. That's not the big 9,000 series, but their 700 or latest 700 series density wise.

9 petabytes of storage in two you of SSDQ of basically QLC storage, which is an amazingly dense package. Uh, rumors have it that IBM is gonna be, uh, announcing the latest version of their LTO drive, which I think is going to come in around 30 terabytes native. So just give, for us old school, uh, uh, storage administrators and people that these numbers are just insane.

The amount of storage you could put into you and the need to be able to archive it is not keeping pace. And it's, it's just an amazing problem that I, I could not have predicted five years ago, let alone, uh, 10, 15 years ago. So it's, it's an an amazing time to be alive and be in storage.

Yeah. So, uh, there was, uh, actually news this week. Uh, Nutanix had their next conference, which I haven't been to since, uh, pre pandemic.

They had in New Orleans, great food, great music. What were the announcements coming out of, uh, next? Well, Nutanix next was in Washington, dc Um, we had a couple people from that were there, and I know that, uh, guy Courier was there.

Um, a couple things that I came out of that one is they did, Nutton has been talking a whole lot about the VMware migrations. You know, they're, they're, they're the alternative or one of the alternatives to Broadcom. Um, and they had clients there talking about those migrations, um, which included people like the US Navy and Moody's.

Um, one of the things I highlighted is that in order, in order to win the Moody's business, they had to have external storage. And so what you're looking at is like, okay, so that's, wait a second. We're, we're, we're integrating all the systems together.

But what clients are saying is like, we've already had this investment in storage, and if you want us to move off of, um, if you want us to be able to move off of VMware, we need to be able to enable some of the storage that we're using. Um, and the second piece of that is, you and I have talked about the POWERFLEX scale. You know, many, it was last year at DTW, we had that deep, you know, sat down and had a deep conversation about Powerflex and, um, with, with Nutanix.

And, and that has finally been released. Um, and maybe you wanna highlight, you know, kind of what your perspective was about why, why they needed to get that out. And, um, it was extremely important to clients that have got, um, very large environments, um, that they need to do a better job of scaling their storage with the servers.

Um, and then this is basically what Nutanix is enabling right now. Yeah. So we've seen this with people who have adopted, whether we're talking about, uh, legacy HPE SimpliVity or their DHCI solution or VxRail, and even in, in, in quote unquote pure Nutanix environments, enterprises don't get rid of their big iron storage array, right?

The, these things are critical for mission, uh, critical deployments and just general workhorses, we're not, we're not even talking about just the, the, the big iron, the, the traditional net fowlers are still critical to mission critical workflows. So I'd said something pretty provocative like seven or eight years ago when Nutanix has said it's kind of, uh, I would say it's hype level, this greatest hype level that HCI doesn't save money. The reason why is because enterprises don't get rid of these legacy technologies.

If you, uh, adopt all HCI, yes, you do reduce your administrative costs, but no one does that. Even in the VxRail environment, you're gonna see plenty of power scale scale. You're gonna see plenty of Power Max alongside of the, uh, VxRail, HCI Nutanix, if they want a shot at these workloads, if they want a shot at disrupting, uh, or taking, 'cause I don't think they need to disrupt it.

They just need to say, yes, VMware, angry VMware customers will, will take your money. If they want to just take this money, then they need a solution that, uh, integrates with overall environments, uh, and is storage and data landscapes, especially when we talk about ai, because these work, the data sets that people are generating from, uh, for their AI is not necessarily landing on HCI. So if we couple the next announcements, uh, to the announcements around, uh, their AI easy button, it all makes sense for Nutanix, right?

And for customers who are looking to displace or have been displaced by VMware's changes in licensing. So it's, it's critically important that they got ahead of this. They did.

I think the timing is well, uh, for them to start to pick up some of those, uh, VMware enterprise customers who, who are not exactly happy with the company. So, on the second, second announcement of the storage piece of it, they announced a, um, relationship with Pure Storage. And, um, they have done some deep integration with their software defined networking that's called Flow.

Um, you know, along with the, what they call NCI Nutanix Cloud Infrastructure, they don't call it HCI anyway. They call it Nutanix Cloud Infrastructure. So our hybrid cloud infrastructure or whatever, let's, let's get rid of that word.

It's kinda like getting rid of data storage. Ah, we're calling it data infrastructure Or, uh, no, like, uh, ai, I mean, uh, development platforms don't like to be called pass anymore. So, um, I was told that, uh, a year or two ago.

So yes, uh, private cloud platforms, Right. There we go. So anyway, so that was, that was another big announcement with, uh, peer storage as being, you know, and so there'll be others, external, external storage companies that will be coming out and, um, probably certifying with this same kind of integration with their, their operating system.

Now on the VMware migrations, um, one of the other companies that was talked about was Toshiba. And Toshiba is migrating, um, two years to migrate 2000 VMs. Comment on that.

That's what they said. Yeah. So the, the, I think just wrote something about the, that this week, uh, that proved to be pretty popular VM to VM migration solved problem.

We were doing this back in 2008, uh, with products like plate spin, VMware, uh, converter, the ability to go from, uh, public cloud, from even private cloud to public cloud and public cloud back from a VM disk format solve problem. Our friends at Veeam do it at, I, I've done it in the CTO advisor hybrid data center and the six five data center where we wanna run a test, uh, sweep on the public cloud or from the public cloud back onto, uh, on premises that's solved. The problem is everything around the, uh, VM and Broadcom knew this exceptionally well when they bought VMware.

The problem isn't, uh, the raw ability to move from VMware hypervisor to KVM Nutanix, A A HV, tho, those, that's routine there, there's tools, plenty of to do that from every provider. The challenge is what happens when I have my security tightly woven into NSX? What happens when I, my storage policies written, you know, hardcoded into vsan?

What happens when I have, when I'm using the integration between VMware, vSphere and my storage rate for administration, when these things don't exist outside of VMware, or not easily mapped one-to-one, you get your two year, you get your two years for 2000 VMs. Yep, yep. And you know, it's, people don't understand.

So something like a, um, data migration, when you're doing a migration from one platform to another platform, it's takes a year of planning. I mean, buddy of mine has had a company, that's all they did. Would they go in and do data migrations for companies?

And he was, he had a team of full-time people, um, back when he worked for, I believe it was at and t, and then he left the left, I think, uh, his D internationals the company, and that is all they do is that migration. Now we've gotten better at it because they have, you know, some seamless migrations where you can plug, you know, plug in the new controller, the new devices, and it'll populate itself. But that's only if it's the same device, you know, so like Pure does that, um, NetApp does that, et cetera.

But if you wanna go to somebody else's system, that's long range planning. And how many times have we seen, this is goes back to my comment that, you know, HCI, the, so I'll say this from seven or eight years ago, HCI seven or eight years ago, this story that you would spend money, uh, to, uh, spend less money. How many times have we seen that you, we see a Hitachi big storage array next to a VM max, next to a net NetApp thing, and this three different generations of boxes.

And it, the plan was always to migrate off one, go to another one. But you get stuck. This, this stuff is complex.

You have low level, your SAP is, uh, directly mapping lungs to the Vmax server. And that, uh, you're doing dis to disc replication from your host. And you can't simply just say, okay, I'm going to plug in another, uh, that abstraction of, I'll just plug in another storage provider behind that.

Just, it's not that simple. It doesn't work that way. And you get stuck in a migration.

And this is something Broadcom has depended on, uh, when it, when it comes to, uh, locking customers into VMware and VCF Broad, uh, hot com promoted that VCF has 70% penetration in their enterprise customers. Yes. Because it's the best licensing option of the licensing options that's presented to 'em.

Customers don't want to change because they have so much other stuff going on. Yeah. Yeah.

Well, enough of that. The m migrations, we've talked about that for now, going on for two years, adding an item. So let's go on to something we don't talk about very often.

That's ai. Yeah, ai, agent ai. We're coming outta RSA, neither you or I cover, uh, security.

So for those of you that are insecurity, we are going to really, really murder what happened at RSA, but it is big news and we do need to cover the news. Uh, basically everything was a gen ai, a gen AI and identity while within security. So this whole idea of how do you secure ai, actually, we, we've gotten briefed by Dale.

We've gotten briefed by all the major OEMs on their strategy around securing the AI data. Pipeline data is something that me and you know about, and this ability to use a, uh, not use ai, but both use AI in, uh, secure ai. The, the, I think the question that was getting answered or asked at RSA consistently, is AI just another workload that needs to be scanned similar to security?

Uh, similar to, uh, how scanners, uh, secure other workloads? Or is it a different approach to security in general? Any thoughts on this at all?

Uh, Kimberly? Well, okay, so a gen ai, um, and identity security, it makes a whole lot of sense to have, um, AI or some sort of agent work on identity access management, I think is what we're talking about here, is are you who you say you are when you're knocking on the door and wanting to access the data? And let's apply those disciplines to there and that, and can we use a ai, uh, tool here to do that work, um, that maybe humans were doing before, or systems were doing before, or can we button it down even further?

So, yes, to me, this makes sense. Um, if that's what we're talking about here, um, I did not attend any of the RSA stuff, so I'm not gonna try to, you know, fake it till I make it here, um, at all. I'll say, I don't know.

Um, but what I do know is that we've built, we, most of the data management companies have start, have started to build or have built some sort of hooks into security systems, um, identity security systems for the purposes of DA data management, especially as we get into the data pipeline. Um, and identifying what, who has what access to what, um, over the time. Yeah.

And then from, as you, uh, connect that to a JA agentic ai, imagine having, uh, to, you know, most people have disabled micros, uh, for the very reason that we're facing with Agen AI security, this idea that there is these autonomous systems accessing your data and doing work on behalf of your organization that can be hacked. So if I can, uh, if you, uh, if I can, uh, do some type of, uh, phishing or something, some social engineering to that agent that has some, some level of intelligence, which can be fooled, how do I protect against that? I don't know the answer to that.

If you want to learn more, go watch the covers from RSA from the rest of the folks here at fu. Alright, so, uh, well, and now I'm going to do my best impersonation of Dion. 'cause Dion covers OpenAI much more than I do.

But there is a big announcement coming out of OpenAI for, uh, uh, their country's initiative. Diane felt this was, uh, exceptionally important. He wrote something on it.

The, I, the general idea is a localized AI for an AI infrastructure for countries. If you're in France, V versus Switzerland versus, uh, the us, you need ai. That's, that's meeting your local re regulatory, illegal and democratic standards.

And OpenAI is specifically going to serve each one of those companies. They announced a $500 billion investment to, uh, to fo to help the US specifically build us, uh, centric ai. Wow, that's big.

That's very, very big. I had not read about that. Um, and this is going beyond the language differences then, where we call it, this is lum beyond low 40 Trump versus of Boots.

Okay. Yes. This is getting into the, um, the norms and the, I would imagine some of the norms that people have as well as the mandates that are coming from the government, which we still are evolving.

And, um, yes, we've got, you know, the EU has put their stuff, has put out their, um, AI guidelines or AI dictates. Um, we have not as a country yet. Um, we've given the guidelines coming from the, uh, presidents, um, and I multiply that because I know we, we had one from the Biden administration, and we're now working on the one coming from the, um, the Trump administration to see what's gonna come out of that piece of it.

But yeah, those things have to be, um, and especially when we get into, I think, well any country, I think that every, we've all experienced that when you go over there, the norms are very, are different. Um, and it's those nuances there that can cause problems with people, um, how languages, how we express ourselves, et cetera. So yes, this, this is big.

This is really big. Yeah. I was shocked when someone from, uh, And is 500 billion enough money for this.

That's kind of my question. Well, The US specifically, the, so, uh, uh, well, I don't know if it's for the US but the, they're talking about building local data centers, local versions of check GPT for the regions that they're gonna operate in. So yeah, 500 billion does not seem like enough.

I think Daniel Newman was just saying, uh, that he, uh, moderated a panel in which the notion of $3 trillion of investment in AI by over the next three years is probably a little conservative. The we're, we're gonna see numbers bigger than that. Yeah.

Yeah. Talking about big investments, uh, less, or are friends at IBM had their big show this week, IBM think or one of their big shows, uh, IBM think is one of the more technical shows, and they announced $150 billion investment in the United States over the next five years talking about ai. So, you know, I think, I, I have to agree you with Daniel, uh, based on these two numbers alone, 500 billion over, uh, I, I, I would assume a shorter period of time for OpenAI, 150 billion from IBM, we've heard the numbers from TSMC, uh, Intel, apple, et cetera.

That $3 trillion number seems a little bit low, but, uh, over the next five years, they're gonna be investing in ai, quantum and, uh, other manufacturing here in the us. And I think the quantum is the big one is because, you know, IBM Manu, what IBM manufactures, they manufacture, um, the storage systems, they manufacture, um, the mainframes and they manufacture the, um, the power systems. This one here is the quantum, the big quantum computing investment that they're gonna be making.

So, um, that will make, and they recognize that, you know, we are well on our way, you know, in March. I mean, it's kind of like talking about quantum computing. It feels like, you know, for 5, 6, 7 years, and we're, we've, it was always seemed like it was 10 years away before it was gonna be ready.

Now we're within a planning period that quantum computing will be ready. And when I say a planning period, it's that three to five year timeframe that we can start thinking about how are you gonna actually really use quantum computing to solve some of the big problems. And one of the things over the past five or 10 years that's changed, that is that Kevin, one of the things that we haven't really thought about is this engagement, this interaction of quantum ai.

And I, I haven't really given it enough thought to, uh, to comment intelligently on it. But as we see quantum start starting to, you know, another non-deterministic technology quantum, uh, uh, take on or integrate with a non-deterministic technology such as LLM and, and ai, how what will the resulting systems look like really intrigues me, really confuses me, and I, I, I, I, it's, it's, it's more technology than I thought I'd see in my lifetime. Yep.

Whole lot. All right. With that, we did our best impersonation of security analysts and, uh, AI analysts from the, uh, from, uh, and quantum analysts.

Uh, but one thing that both me and Kimberly know is storage and systems, and I think we covered that exceptionally well. So a plus on storage and systems, c plus on everything else. If you wanna learn more, uh, tune in next week.

'cause I, I'm asserting our good friend Diane will be back to help pick up the slack and some of this conversation around compliance CIO level stuff. Remember to follow his, uh, CIO checks, CIO, uh, hashtag CIO checks every Thursday. And, uh, follow my zero to builder series, hashtag zero to builder to learn how to develop code from zero to build in a hundred days, even if you haven't joined.

It's a, it's a good opportunity to, uh, catch up. Thanks everyone. Have a great week.