Clips
News
More from VMware Explore 2024

Brock Mowry

CTO

Tintri

Phil Trickovic

SVP

Tintri

Kubernetes evolution: Tintri’s strategies for enhancing cloud-native architectures

With the fast-paced advances in technology, the challenges and opportunities surrounding cloud-native architectures and Kubernetes management are more pronounced than ever.The shift toward decoupling traditional three-tier architectures to embrace more agile, scalable solutions such as Kubernetes is no longer just a theoretical conversation. It’s becoming a practical necessity for enterprises across the globe, according to Phil Trickovic (pictured, right), senior vice president of revenue at Tintri by DDN Inc.Tintri’s Brock Mowry and Phil Trickovic talk with theCUBE about Kubernetes management

play_circle_outline Tintri's focus on QoS and controlling lanes for applications

play_circle_outline Tintri's CSI driver for generic Kubernetes and data migration

play_circle_outline Simplifying operations and managing exponentially growing data sets

Info
Transcript

Phil Trickovic & Brock Mowry | Tintri

Brock Mowry

CTO Tintri

Phil Trickovic

SVP Tintri

At VMware Explorer 2024, SVP Phil Trickovic and CTO Brock Mowry of Tintri discuss Kubernetes and its impact on cloud native microservices and applications. Tintri focuses on providing QoS and control for these applications through their platform. They have enhanced support for various hypervisors and containers, including Tanzu and generic Kubernetes via CSI, allowing for snapshots and data migration capabilities. Tintri's intelligence at the IO level aids in optimizing workloads and managing data sets efficiently. With the acquisition by DDN, Tintri compleme... Read more

explore Keep Exploring

What has Tintri been focused on for the last three years? add

What capabilities does the CSI driver created by the team offer and how does it leverage the power of Tintri for app migration and data management within Kubernetes environments? add

What was the foundational principle behind Tintri's creation and how does it relate to the architecture of VMware and processor wastage? add

bolt Powered by CUBE AI

Phil Trickovic & Brock Mowry | Tintri

>> Welcome back everyone to theCUBE's live coverage here in Las Vegas for VMware Explorer 2024. I'm John Furrier, host of theCUBE with my co-host, Dave Vellante. Two sets of coverage. Phil Trickovic is here as SVP and Brock Mowry is the CTO of Tintri. We've covered many times on theCUBE, now back and continuing to stay in the game, now part of DDN. Gentlemen, thanks for joining us on theCUBE here live at VMware Explorer.

Phil Trickovic

>> Thanks for having us.

Brock Mowry

>> Thanks for having us.

>> So I love this session. I was looking forward to when I saw it on the calendar. Kubernetes our favorite topic, of course, we were there. President at Creation at the original KubeCon when Linux Foundation established the CNCF, and it's an ecosystem and it's really powering cloud native. Kubernetes kind of getting boring, which means it's working and cloud native microservices are now... Containers are now powering this next gen foundational system. So there's a lot of conversations like, "Hey, forget the dogma. I just need the foundation bill so I can get Gen AI data models up there." So there's a lot of pressure in that community to say, "Don't fight those dogmatic wars around this, that kind of thing, just give me Kubernetes." Okay. Running at scale, this is kind of what you guys are doing. Give us an update on Tintri.

Phil Trickovic

>> Yeah, so from the business front, you've nailed it in that when we're seeing this across the stack from biggest enterprise down through mid-size, even to the smalls are investigating the possibility of Kubernetes deeply. But what they're getting to a point of is that the non-cloud infrastructure or platform entropy that we're experiencing with these three tier architectures that we are been bound to since 1990, do not allow you to progress into those applications in any way that makes sense financially or talent pool wise. And yes, that's what Tintri's been focused on and what Brock's been focused on for the last three years is getting our QoS and ability to control those lanes exposed to those applications and move that forward.

Brock Mowry

>> Yeah, I mean, one of the things that Tintri's always been known for is its intelligence of the workload that's running over top of it. We were trying to bring that same intelligence into the Kubernetes space, being able to show performance and latency information at the PV level within the Kubernetes environment, and that's something that we don't see out in the environment.

Dave Vellante

>> Our friend Jeff Jonas says, "Entropy's winning."

Phil Trickovic

>> It's absolutely winning.

Dave Vellante

>> So how do you balance the score?

Phil Trickovic

>> That's the unknown. That's part of the either... Whichever side of the entropy scale you want to be on, you want raise it or lower it, it's not necessarily good or bad either way. Right now, our industry and our globe is stuck on an architecture that was designed and started to be architected in the 80s. It is. Still to this day, three tier plus cloud. So three tier plus ASP, three tier plus whatever. Going back, how we're doing that is decoupling that stack. So our intelligence goes from the application down all the way through the storage into the backup if you want to monitor that as well. All statistics on the app, all performance stats, all localities etc. that you've got with VMware on us is being extended into those workloads. That's important because with VMware, any x86 architecture, you've got NVMe as an option. You've got SCSI bus fire or SCSI fiber channel etc. The limitations on the command queuing for those is going to prevent this from ever happening because even on NVMe where you're at 68,000 or whatever it is, streams, you're talking 1.2 billion, sometimes streams or zillions of points that have to be processed to deliver a result. Cannot happen the way that... And I'm talking globally is architected at the energy conversation.

Phil Trickovic

>> Before we get into some of the conversations, I want us just reset on Tintri. We had an old saying, I'm dating myself when I say this. When something changes, you say, "I want to show the new Polaroid picture of what I am or what we are." Now, what's the new Instagram photo of Tintri? Give us the update because we covered Tintri great product leadership at a time where that market was surging, obviously from storage and changing with the hyper-scalers and then hyper-convergence. Now as part of DDN acquiring you guys, you're back on the front lines again with the product market fit happening. Give us the update. What's the new Instagram of Tintri?

Phil Trickovic

>> Well, a little because it's more on the technical side from the company.

Brock Mowry

>> We wanted to take a focus, again, back to the core product and then expand it out to be able to support multiple environments. It just so happened there was a hypervisor shakeup here recently. Lots of conversation around it. Prior to that happening, we already started looking at this saying, "Hey, we need to expand our support for these other hyper-visors. We have very deep support with Hyper-V, and a couple other hyper-visors, but there was some other ones that people were asking about key support in that area. So we've bolstered our support in that capability, but then we bring in the container support. We did a round with Tanzu. Our most recent here that we're talking about now is generic Kubernetes via CSI, and we're starting to get into a lot of great conversations.

>> Those are engineered solutions with Tanzu?

Brock Mowry

>> Well, the original one was with Tanzu. Now this announcement with our CSI now is actually generic community Kubernetes through CSI.

Brock Mowry

>> Talk about that announcement real quick. Give an update on that.

Brock Mowry

>> We're in the final stages of this release, but we've basically been able to create a CSI driver that can consume storage off of our platform, do all the provisioning, everything like any Kubernetes admin would expect. But we also bring the power of Tintri. The QoS, like I mentioned before, the snapshots. Some of the other things that I've challenged my engineering team was, look, we've got both of these environments sitting on top of us. Why does somebody need to duplicate or move data to do an app migration or something like that? So we're looking at the ability to snapshot data out of a virtual machine and populate a persistent volume. That could be huge for companies that are trying to move large data sets in and out of Kubernetes environments.

Brock Mowry

>> Like who? Give an example of a customer and what their problem statement would be for that.

Brock Mowry

>> We've seen some customers that... We'll use the AI model as an example. You generate some of these models and you've got these large data sets. Well, at some point you need to duplicate that data set so other teams can do other things with it. VMstore shines there. We can do a zero capacity impact clone of that data as well as protect both the IOs to that data set and be able to provide the ability for two different teams to interact with the same data, but have it separated.

Dave Vellante

>> If I recall, so Tintri's innovation early on was you guys were really focused on simplicity and kind of beat the market back when it was before the cartel got a hold of all the VASA APIs and then-

>> Cartel.

Dave Vellante

>> Well, it's true. That's what happened. So you guys were ahead of the market and then all of a sudden it's sort of the API integration commoditized everything. It sounds like you're now taking that concept up the stack and driving new value.

Phil Trickovic

>> Yes, because you guys have been with us for a long time and covered us since the beginning, pretty much. The foundational principle of Tintri was to eliminate a huge waste back that was generated basically the way VMware was architected with the way the IOQs were happening to the storage, processor wastage. Why did VMware come to be?

>> .

Phil Trickovic

>> We had the Pentium Pros. You guys know this, in 2000, we haven't improved our processor speeds much, like incremental and the energy suck. The amount of greenhouse gases we pump in to get a 10th of a percent in the gigahertz is completely unnecessary. Same was true when Tintri was founded, when VMware came out and reduced that and started using those processors more. So where are we now? These processors in the infrastructure cannot handle it without that queuing that happens on the backend. It cannot. There's too many queues, and that's what our specialty is, is the intelligence behind those data sets.

Brock Mowry

>> And acceleration too.

Phil Trickovic

>> Paramount of success.

>> It's funny, in GDC, NVIDIA is banking all accelerated computing. I mean, all these innovations around the system now, not just one thing.

Phil Trickovic

>> No.

>> And so you're starting to see this clustered system become super computing, which puts more pressure on the cloud native community because now, hey, the horsepower's here and the distributed systems now were working well, let's get the Kubernetes layer to orchestrate. This is now pressure on KubeCon. I mean, KubeCon's been doing great, but it feels like it's had a moment where, "Hey guys, we just got to unify this. Just no more discussion. Here's what it is." And VMware's very much in the mix of this too. Do I run VMware in a container or do I run it with Tanzu? So what's your take on this? Because this is the forcing function that it's almost like if this doesn't happen, we can't go to the next level.

Phil Trickovic

>> Well, we keep going... Okay, I'm going to digress on this in a minute. Have you guys covered Grok or Cerebras?

>> Absolutely.

Phil Trickovic

>> How about Cerebras?

>> Just their news this morning on the-

Dave Vellante

>> Cerebras. Yeah.

Phil Trickovic

>> Cerebras, whatever.

>> Cerebras, their inference engine they launched today and NVIDIA click killer.

Phil Trickovic

>> It's fricking amazing.

>> Yes.

Phil Trickovic

>> Those guys pumped out 1.2 trillion transistors on a... It's a big platter, right? What's the issue from what most enterprises though? Is they're not going to be able to afford these platters that do this, those are going to train the model. So we're at an inflection point in the market that we both think happened around 2019, where you've seen this training craze and it's amazing what we're going to solve with this. This is a good thing. It's not bad. They've got all the efficiencies that they know what they're doing to spit these models out. When you try to pour it over to enterprise, where you don't have 50 PhDs who are familiar, who have wrote... Whatever the situation is, managing and deploying these systems like the talent pool, the brain trust, the deliverance not there yet. So we simplify where you don't have to worry about any of that, just like we did with VMware, Citrix .

Dave Vellante

>> The parallels are amazing.

Phil Trickovic

>> Yeah, it is.

Dave Vellante

>> I used to say these guys in lab coats to figure out what happened at the IO level if something went wrong.

Phil Trickovic

>> It's serendipity.

Brock Mowry

>> We need the intelligence at the IO level.

Phil Trickovic

>> It's serendipity.

Brock Mowry

>> And then be able to provide that information back to the infrastructure to automatically adjust and optimize.

>> One of the areas you guys are on that's interesting is that DevOps now DevSecOps, it's about operations. VMware's in that ops world. You guys are in that how to make operations better. And now that data is becoming part of the developer equation. We see that the data engineering is going on with platform engineering. So you're seeing a nexus of those two worlds coming in. These aren't analytics data scientists. We talking about coders, system coders doing data engineering. So you got this overlap in platform engineering and data engineering emerging. One, thoughts on that and two, how do you operationalize this?

Phil Trickovic

>> God, man.

>> How much time we got?

Phil Trickovic

>> two days. I could cover the developer-

Phil Trickovic

>> Brock, we need an algorithm done by the end of the interview.

Phil Trickovic

>> And everybody who's in tech has experienced this from one side to the other. DBAs, developers, infrastructure, your fault, your fault, your fault, your fault, your fault, your fault.

>> And they're all right.

Phil Trickovic

>> That's been the problem, unless you're on Tintri where we can tell you exactly where that is falling down, and that's where the intelligence is critical because these workloads and these data sets are exponentially going to explode. It's not going to get easier to manage this stuff. It's going to get more difficult if you don't have intelligence.

>> So manageability is the key.

Phil Trickovic

>> Intelligence of the data sets key. Even before, I would say, me before... Management's key, you have to know what you're trying to accomplish. What was it? What was the paper where it was like a zettabyte output file to run whatever it was running? Who's going to manage that? What are you going to do? What do you do when you have to shard that out to get recency of data? All of these things require intelligence that the enterprise do not have.

Dave Vellante

>> So where's that intelligence come from? What's the sort of secret sauce underneath?

Brock Mowry

>> Well, so VMstore has traditionally been a product that has incorporated a form of machine learning, artificial intelligence, and it manages those IOQs. It optimizes that workload across our appliance entirely. That is one area that we see is going to help that industry. The other one is other technologies that help bypass some of these or eliminating layers. You look at the AI GPU architecture right now, you've got x86 bypass with direct GPU storage. They're going to have to come up with other types of architectures that get past that x86 barrier that allow the front end and the backend to work simultaneously and extremely fast.

Dave Vellante

>> Because the x86 is controlling everything, right? You're saying-

Brock Mowry

>> Right now it is. It's kind a roadblock.

Dave Vellante

>> Sp all the x86 memory. It's the bottleneck.

Brock Mowry

>> Yep.

>> We just did a super cloud seven on the future data platform where we were kind of saying the modern data stack is dead. Because you look at companies like Uber, you have a lot of heterogeneous things on data store, and then you have all these new system architectures around chips and gear, compute networking and storage. What is an intelligent data platform look like as you guys see it? Because this is what I see happening, and you guys are in the middle of it. What does a company do to invest in a generational foundation to get there? Because we hear from customers, "Hey, we're built for the next 10 years." We want to have a solid foundation. That's why the Kubernetes discussion... Solve it. Just get it up and running. I want to build on top. What's your thoughts? Brock, we'll start with you.

Brock Mowry

>> I'm blanking. I lost it. Sorry.

Phil Trickovic

>> Okay. I want to bring up a project. Brock and I have been involved in some really large projects, one with one of the biggest energy companies in the country. This is five years ago?

Brock Mowry

>> Yeah.

Phil Trickovic

>> They were converting the... I don't know how much of this I can say on... Can you guys edit this or is this?

Phil Trickovic

>> No, it's live.

Phil Trickovic

>> We're live. Nothing but a live show-

Phil Trickovic

>> You can talk in generalities.

Phil Trickovic

>> Generalities. So one of the... The company that basically runs most of our nation's power grid containerized back in what? Started 2018, 2019.

Brock Mowry

>> Yeah, they were early adoption.

Phil Trickovic

>> 175 people to do this conversion. So it was very difficult back then. It was very resource intensive and require a lot of skill and a lot of oopses and a lot of . But now that's the standard. And what that's done is it decouples all the server requirements, a lot of the network requirements and a lot of the end user nonsense that you have to go to. So even with that staff, the ROI that seats you on, it still made sense. And the 10-year one's insane. It's an order of magnitude reduction in cost and management once you get it stood up. So my advice to businesses that are now starting this journey, just start the journey. Pick one application to start, call experts, get us involved and let's start looking at it. What would-

>> Decomposing it and decoupling it was the strategy there that worked for them.

Phil Trickovic

>> Yeah. But it was an army to get it done. And again, this is a massive scale. I don't even know what the budget was, but it was a huge project. Greatly successful though. And the return on it was insane what they saw when it was done.

Dave Vellante

>> It's interesting. I want to pick up on something John was saying about the new data stack because we've written, beginning to think a lot about this, where everybody's talking about agents as a co-pilot, but not a lot of people are talking about multiple swarms of agents working together. And to make that work, you're going to have to tap into some of the backend systems and get to the data. What I understand, if you can containerize that and make it easy for the up the stack application developers to get access to that data, then it can be harmonized. That's not your job to harmonize it. So that revenue means revenue. So that we're not in a meeting arguing about, "Well, what's that mean? Is that bookings or revenue?" And I mean, it means the same thing. So that's somebody else's problem. But you guys can help connect to that data and surface it so that it can be harmonized and then agents can actually operate. It can be governed. And again, that's not your purview, but the starting point is that infrastructure to surface that data.

Brock Mowry

>> Right. Absolutely.

Phil Trickovic

>> You nailed it. And it's simply done versus having to go and write a script that would go from the ceiling of the floor on how to access a set of data blocks. We could turn that on tomorrow.

Dave Vellante

>> So you think about John's example of Uber. You got riders, drivers, ETAs, destinations, transactions, dollars. They're all different data elements-

Dave Vellante

>> And companies.

Dave Vellante

>> That sit inside of different systems. And Uber had to write the 170 people, like the energy company you talked about, they had armies of engineers who had to develop that. We think the future is Uber for all, right? But the backend has to get solved before any of that works.

Brock Mowry

>> Yes.

Phil Trickovic

>> Big time. Because that backend process that you're speaking of is exponentially growing by the week because we keep adding new data sets to it. And Brock was saying something earlier that's really important is the criticality of being able to manage and recover those data sets in their original state is going to be first and foremost. Because if somebody gets in and changes something that you're running a trained model against, you poisoned that model and that entire effort is out the window. It's incredibly difficult to track that now without something like what we do.

Dave Vellante

>> How is the DDN DNA, the HPC DNA, which you've talked a lot about-

Dave Vellante

>> They're at GTC.

Dave Vellante

>> Coming together with AI. I mean it just seems like there's a huge tailwind for DDN. How has that DNA affected you guys? Where do you guys fit in that DDN portfolio?

Phil Trickovic

>> Do you remember the split I was talking about? That's what's perfect about the role of the relationship. They are the masters of that world. I don't go there. We don't go into that space, but they own it and they're fantastic at it and they support what we're doing and we own the enterprise space, the deployment space. So the marriage of the two is like-

Dave Vellante

>> It's interesting because they've always wanted to get into that space and they try to take their novel high performance architecture into the enterprise. It didn't work. So the acquisition of the IP that they picked up from Tintri has enabled them to get into that market. And they made a couple of other acquisitions too, I believe.

Phil Trickovic

>> Yeah, they have. They do on occasion. I think the focus for us now, like DDN, speaking on a bigger picture for DDN is exactly what we said. They're going to own that space and they are getting there with NVIDIA as partner and all the good stuff for the training dimension of it.

Brock Mowry

>> And these data platforms are emerging too. It has all the storage, all the infrastructure underneath, and there's all that kind of squishy stuff that you guys do that's in there. It's a nice little bolt on with DDN.

Phil Trickovic

>> Yeah, no, it really is. It's very, very smart of them to-

>> AI's got some good business for you guys now, you can smash with some AI business.

Phil Trickovic

>> Yeah.

Brock Mowry

>> There's also a bright spot for the VMstore as well, because when you getting into the lower end enterprise AI, if you will, VMstore shines when you have multiple science projects going on on top of it because of that quality of service at the -

>> You're enabling value on top of you.

Brock Mowry

>> But then we complement the DDN side because we don't step on their territory, they don't step in ours. They are the big iron, we are the training environment or the testing environment.

Brock Mowry

>> You give flexibility to the customer to architect whatever they want.

Phil Trickovic

>> That's what's beautiful. We've got the one-stop shop for it all the way up through the app stack up to the network.

>> One-two punch.

Phil Trickovic

>> And a lot of expertise.

Dave Vellante

>> How's Alex doing?

Phil Trickovic

>> Doing great.

Dave Vellante

>> Did he help you buy that suit? Because man, he's got taste. You know what I'm talking about. He's unbelievable.

Brock Mowry

>> .

Phil Trickovic

>> a stylish guy.

Dave Vellante

>> Best dressed in IT and tech, he is.

Phil Trickovic

>> He really is. .

>> No doubt. And we get Joseph George over there now from HPE.

Brock Mowry

>> Yes.

Dave Vellante

>> Kube alumni, Joseph George. So he just came over. Anyway, guys, great to have you on. Final question. Give a plug for Tintri real quick while you're here. I know you got all new stuff up. It's still the same game, but new ball game on the field. What's new? Put a plug in for Tintri.

Phil Trickovic

>> What's new or give you a Tintri plug?

Dave Vellante

>> Both.

Phil Trickovic

>> All right. Just call us for the intelligent aspect if you want to get into the rapidly being adopted space of containers, we are the easiest path to do that.

Brock Mowry

>> All right. Easy button for data management. That's really what it comes down to.

Phil Trickovic

>> That's a bumper sticker right there.

Phil Trickovic

>> We're all about the data. We don't care about the hardware-

>> Love that bumper sticker. That's a New York Post headline. All right guys, thanks for coming on.

Phil Trickovic

>> Thanks, guys.

>> Okay, we'll be back with more coverage. We'll be right back with our next guest. I'm John Furrier with Dave Vellante, host of theCUBE. Thanks for watching.

Dave Vellante

>> Thank you.