Facemelting w/Chad Sakac Pt.1 - The CTO Advisor Podcast 088

Machine-generated from the episode audio and not hand-corrected, so names and technical terms may be imperfect. The audio is authoritative.

Hey, how's it going? It's Keith Townsend from the CTO Advisor, and you're listening to episode 88 of the CTO Advisor podcast. On the line, I've actually went out onto the Twitters and found Mark May. Mark, how's it going? Good. I do exist. I am still here. I've just been busy, man. I haven't seen you in a while. Yeah, man. The rumors on the Twitters is that, you know, we had Greg Farrell on last week. The rumors on the Twitter is that you guys just can't get along.

Oh, you know, Greg, everybody just wants to punch him in the face and hug him at the same time. Yeah, he's kind of like, I'm coming off of HPE Discover, and they had the cutest mascots. There was the IT monster, and you wanted to hug him but punch him in the face at the same time. I saw those. I need to get me some of those since I didn't make it out. You guys have got issues to work on. Get your own shoes.

All right, and that voice, you're not mistaken. That voice on the other end is Chad Sackett. Chad, how you doing? How's it going, man? Life is grand, and Mark, since he's storage, it's good to talk to you guys both. This is a long time coming. We've been trying to record forever. Well, well over a year. It's not a lack of will. It was the gods of the calendar. That's how it goes. So it took – I had to make a special call to Michael.

I said, Michael, I don't know what it is you have going on with Chad, but I can't get him on the podcast, man. There's going to have to be a – something's going to have to give here. You've got to get him in a role where he has time to podcast. Right. He needs to be able to talk to the people. Chad is a man of people, and he needs to talk to the people. Speaking of that, lots of HPE for you, Ted.

Yeah, it has been. I guess I can now kind of imagine how that HPE mascot would feel. At the same time. Yeah, no, it's been a lot of change over the last couple of years as a whole actually, Mark. You know what I mean? Yeah. It's a new gig, and that's exciting, but it's been an interesting couple of years, not just obviously industry-wise, culture-wise, what's going on the world-wise, but in little old Chad land too. So, Chad, give us some hero numbers, man.

That what was VCE business that you morphed into the – I don't know if it was the converse business or whatever it was in Dale. That grew from what to what when you left it? Like, let's do a little chest thumping. So, I started leading that group in the beginning, like literally January 1st, 2016, and it was a business that did about $3 billion in earnings when I took it over. That was like a run rate versus an in-year number. 4.

So, it grew a lot over that time period. The thing that I was more proud of the results in the team was it was also a huge transition period. So, the CI business, like if you imagine coming into it, there was a huge VX block business, but there was a nonexistent HCI business. And that HCI business basically grew to a billion and change on its own. 2 billion just as a standalone business at the end. 2 billion in that whole net new part of the business in the span of two years, which we knew that there was going to be more and more growth in HCI than there was in CI.

And it was tricky, right? Because if you think about the brand of VCE, what do you visualize? The V block? The V block, yeah. And rightfully so, they're awesome and customers love them and all that jazz. But whenever you have something that's, you know, it's surprisingly hard to create change when something is good. Right? When there's no crisis. Right. And so, you know, what basically had happened was we kind of knew even before the Dell EMC merger that HCI was going to be really big and we needed to run really fast to catch up to Nutanix and to others.

And the team managed to do it, which was a big accomplishment. So, to put some context behind this, I interviewed you in the other chat during Dell Technologies World 2018. Yep. And Chad Don put out a number out there that was really interesting. 62% of the customer meetings that week was about HCI. And, you know, I tweeted that and people were like, oh, yeah, so what, 62% of 60 meetings is not a lot of meetings. So, to put proper, I think I got some clarification.

That was over 200 customer meetings was about HCI. Mark, I talked to a customer, I think the very same week, who was a huge SAP house, and they were looking at migrating to HANA. They were coming off of HP UX and Itanium servers. So, they're looking for a modern infrastructure. And there's this whole big thing with SAP and the rest of the industry about certifying HCI or HCI for HANA. This customer still chose VxRack, VxRail over traditional converged infrastructure systems because they wanted that architecture that much.

And they were counting on that, you know, within the next few months or a year, SAP will get its act together and certify these systems for HCI. To follow on to that, I interviewed, I was at SAP Sapphire a couple of weeks ago, interviewed the alliance partner who's in charge of, you know, kind of these HCI relationships. And he said that SAP is doing HCI, HCI, not because SAP wants HCI, but customers want HCI. So, it's an amazing change in the past year or two.

Well, if you think of what customers got with converged infrastructure, the simplicity, the ease of use, we get all of that plus a different scale, more agility, and a whole bunch of other benefits with HCI. I mean, it's, and I think generally speaking, it's probably cheaper. So, there's no reason why customers wouldn't be clamoring for something better or at least easier. Yeah, you know, the thing that was, I want to be, there's always a tendency in our industry to, you know, have hyperbole kind of make, you know, given A, B, C, immediately our brains go, oh, that's a multiple choice question, pick A, B, or C.

You know, and it's generally more along the lines of, okay, it's really how much of A, how much of B, and how much of C, and how are they going to grow and change over time? The CI business continues to do okay, but the growth has slowed and it's been replaced by super high growth in HCI. And as nerds and proud nerds and technologists, we look at them and we focus on their differences. But, you know, Mark, you hit the nail on the head.

They both are in essence, hey, this is something you consume, you don't construct it. And somebody else takes the accountability for design, lifecycle engineering, sustain, support, all of those things for the stack, right? And HCI and CI, if you squint your eyes and step back enough and you're not looking at how they're designed, they're the same core value proposition, right? But I can tell you this, the person who is the person where the buck stopped for many, many thousands of customers, when we do those platform updates, and any VxBlock customer knows this, when we do an RCM, which is a release certification matrix update, which means that we update everything inside the VxBlock.

It takes months of planning, like it takes a lot of work. It takes a lot of like check, check, check. And then the team at Dell EMC do a really good job of doing those platform updates. The interesting thing that was the flip side is that the first time that we had a major update to the VxRail base, and we needed to literally update every single VxRail that was out there. U1 to U2, right? 2 update. We had to go out and touch the entire user base.

And literally within a period of a few weeks, we had gotten to like 90%, meaning that the main difference between CI and HCI is, yep, one can start small and scales differently. The other one kind of starts big and is designed and deployed in blocks. But the bigger difference is that the process of doing a lifecycle update when it's software as opposed to, you know, a hardware stack is a lot easier to automate. And, you know, again, I don't want to get hung up on it, but it's interesting.

The one thing that I'd say as I, you know, as I age a little bit and get a little more experience and do different things in my career, you start to realize common patterns. One common pattern that I see that is across domains, whether it's infrastructure or even the new domain of Pivotal, is this debate over do you build it or do you buy it? Do you construct it or do you consume it? It exists everywhere. You know what I mean?

So for those of you wanting to know the Pivotal story, we'll get to that. But we're going to save that for the next episode. We're going to split this conversation into two parts. I want to focus this conversation around converged infrastructure, HCI. For sure, man. In general. But help us out here with this concept of CI. Dell EMC came out with a new Vblock. And initially you look at it, you scratch your head, you talk about the rocket ship that is the VxRail, VxRack thing.

You know, you talk about the 62% customer meetings. What's the market for the traditional CI? I'm sure there's a whole bunch of Vblocks or VxBlocks out there. So there are thousands of VxBlocks out there. So you've got to visualize that CI market, Keith, is call it roughly around a $5 billion to $6 billion market. It continues to grow at 10%, 15% per annum, which is higher than the growth rate that you see in infrastructure as a whole. It's dwarfed by the triple digit growth rates of HCI, but it's a big business.

And the long and the short of it is that there still isn't, as far as I know, an official update of the SAP HANA TDI test harness. No, there isn't. You know, for HCI. And that does not mean that it does not work, right, just to be clear. Like that test harness, what it does is it tests the performance of log writes and commits and cluster failures. And all that stuff works fine on HCI. But, you know, A, some customers want traditional storage stacks for either things like certification.

So another example is that for EPIC, which is, you know, a very common EMR, electronic medical records application stack, and particularly in America, but also in Europe, portions of EPIC can run on HCI, but portions are not validated to run on HCI. Then there are some things where the workload itself, not just because of certification, but for real technical reasons, needs a specific data service. So an example of that is that if you had a workload that needed to have very, very high rates of synchronous replication, none of the HCI software defined storage stacks can do that particularly well yet.

Right. So that's a driver. And then the last one, and if anybody doesn't trust me on this, just do the math. I've done the math. This is not an opinion. This is science. Depending on certain workloads, CPU and storage scaling rates, in some cases, CI can be materially less capital intensive, even though it starts much bigger. You know, at certain scale points, particularly places where CPU and persistence scale very separately, it's architecturally lower CapEx costs. Forget about the OpEx for a second, like how simple it is to run an update.

But the capital cost between CI and HCI can vary. So there's three reasons. One, safety and certification, you know. Two, some workloads really need traditional data services for reals. And three, certain workloads, again, scaling and CPU storage ratios fit CI better. An example of that, by the way, is that, like, a lot of people deploy end-user computing workloads on HCI because it's operationally easy and it's easy to scale up HCI as you go through pilots and then deployments and you grow your user accounts.

The flip side is if you're a customer that says, I know I'm going to deploy, like, 12,000, you know, Horizon instances, in some cases, it can be a lot less capital intensive to deploy it on a VX block with ExtremeIO, as an example. Mark, you do a lot of large-scale deployments in general. Does that kind of represent what you see? Yeah, I mean, it definitely does. More and more people are – and this is – we've been going down this path for a while of trying to focus on business value.

And, you know, something I've said many times over is there's no business value in running cables, right? That's something we have to do or has to be done to enable business value. But the more we can get out of that, the better. I think that was the fundamental selling point for me of any type of converged infrastructure. And it just gets better and better as we move down the HCI path. And I think to circle back to what you said about SAP and qualifications, sometimes you just don't care, right?

Sometimes we know that certain vendors, you know, they're just behind the curve and you end up doing what you feel like you need to do because it's right for your business and it's right for your environment, with or without that certification. I mean, just think of when VMware was new, right? Oh, if it doesn't work, we're going to not support you because you're not on VMware, right? That mentality goes away eventually. I think that same thing happens with HCI. It's quicker and there's people that are struggling.

You guys have been around the block. We have all done it together. In fact, our paths first crossed in the early days of the VMware revolution, right? We've got to put ourselves back in time then. I don't know if you remember this. In the early days of ESX existence, people dug in their heels and basically said it's not going to work. They have no basis to say it, but the default position was danger. Danger will rob it, right?

I find a lot of that is people just reacting to fear of change, fear of the unknown, feel like, oh, my God, the sky is falling. My job is going to go away. But I think that's just completely unfounded on all accounts, of course. So let's talk about the software components of CI, HCI. I want to pull up an interesting tweet from earlier today. CNBC put out that, obviously, for various reasons, Intel is looking for a new CEO. Well, we'll leave that to you guys to do a little Internet research on why the Intel CEO got ousted.

But logically, you look at it, you look at the industry, you thought, OK, who's the next perfect CEO for Intel? You look at this little company called VMware and you have in VMware this Intel alum who's grown VMware to be a pretty substantial business. And the rumor gets put out. You know what? Pat should become the next VMware Intel CEO. Pat quickly replied, you know what? Thanks for the mention. But I love my job at VMware. Software is the future.

So talk to the importance of software when it comes to CI and HCI. You know, Nutanix has famously become a you know, they gave the they've modified their OEM relationship. I think it's super micro. And now they don't sell hardware directly anymore. They're a software company. So speak to kind of the importance of software when it comes to CI and HCI. Yeah. So so first things first, there's there's one thing that's common in the software stack for CI and HCI.

And then there's one thing that's different. The first thing that's the common element is that the software is the glue that brings together the system. So particularly when it comes to the control plane. So control plane meaning, you know, not the data path, but how do you interact with the thing? This is an area where we made progress on the VX block business while I was there. But I'm glad to say that, you know, the new team that's continuing to lead the charge is making even more progress.

How do you make it so that the system has got a common API, that telemetry looks and feels a certain way? Ultimately, whether it's manifested in CI form or HCI form, you should be able to do a single a single command or single trigger event driven platform update. Right. So the software has to go through the lifecycle management of the stack. So in essence, management and orchestration is a critical software component in both CI and HCI. Now, in the HCI case, the second critical component, which is that that is actually architecturally similar in both CI and HCI.

Not the same, but functions exist in HCI. In HCI, there's another critical feature, which is one of the defining elements of how the system behaves, which is the software defined storage stack. Right. Because in the case of HCI, there's no array. You're basically using some form of software distributed storage. And if you take a look at the HCI vendors that have come and those that have gone, one thing that is a common element in those that didn't make it is, in many cases, their SDS was not particularly good.

And if the SDS is not particularly good, the difficulty is, is that the SDS layer A defines how does the system deploy? How does it scale? What data services does it have? But the most important thing is anyone who's been in IT for more than a little while knows that the worst day you can have in IT is a bad storage. They are the worst days. Right. And the reason that it's the worst day is storage out of the three primary domains of IT, compute, network and persistence.

Storage is the only one that actually persists information. You know, and the easy way to think about that, right, is like if a host gets cooked, right, and you have your storage on a network device. OK, no problem. Right. Anyone who has a computer knows. OK, great. I'm just going to replace the server. I'll remount the network mount and I'm off to the races. Right. If your host catches fire and the storage is inside the local, you know, your local computer at home.

Right. The thing that's going to take the longest to recover from is not replacing the hardware, it's replacing the data. And therefore, in HCI land, you know, if you take a look at the two really dominant HCI players now, it's basically VMware and Dell EMC and then Nutanix. They have at their core, you know, a software defined storage stack, vSAN in the case of VMware and Dell EMC and the base MDFS layer in Nutanix that are time tested and true storage stacks.

Building a storage stack is hard. Yeah, I remember Pat at VMworld, maybe it was 2015. And they were talking about vSAN and, you know, the pace of implementation when it came to vSAN. They said, you know, storage is hard. He said he's designed processors and getting storage right is is really, really hard. So we're going to hit the pause button. That was part one of our face melting conversation with Chad Sackett. And part two, we're going to talk the pivot to full software.

We're going to continue this software conversation. We're going to pivot. Yeah, we're going to pivot. We're going to pivot and talk about Pivotal. Chad, for those who don't know where to find you, where can they find you? You can find me on the Twitters at SACKACC, S-A-C-K-A-C-C. io. And Mark? You can find me on the Twitters at SensiStorage. All right. com, on Twitter at CTO Advisor. And as always, I haven't said this in a while, so your mom may be wondering what happened to their her CTO dose.

Update our podcaster, the letter, you know, hear the greatness. That is the CTO Advisor podcast. Share this with your friends. You can find it. Find us in iTunes and your favorite podcaster.