Transcript
Hi Mike Matchett Small World Big Data and I'm here today talking about one of my favorite kinds of technologies, which is storage that's tailored for containers. I've got tintri here. We've covered Tintri. I've covered Tintri for many, many years, um, uniquely built for each VM. And you could instead of looking at big file volumes and everything else, you could deal with that once the plastic containers. So just stay tuned. Oh. Hey, Brock, welcome to our show. Thanks for having me excited. Uh, so you're the CTO at Tintri. Um, and, and you get to look at a lot of cool technologies and things like that. It's it seemed like, uh, you know, what we're going to talk about here is a natural fit for VM store, which is to say, let's make bespoke storage that you can manage at the same level that you manage the things you're managing in application. So whether it's a VM or a Kubernetes cluster or a container. We'll talk about that. But before we get into that, could you just give us a thumbnail sketch of sort of the general Tintri VM store value proposition for folks? Like what? How would you describe the sort of the unique technology that you guys bring to bring to the to the VM world or the storage at large? So, you know, one of the one of the big advantages that the VM store brings is extremely powerful capability, visibility, performance and so forth with very little administrative overhead. So care and feeding of this product is is very minimal, which is, you know, a huge challenge, especially when you start getting these platforms up to a large scale. The more efficient or the more autonomous they can operate, obviously the bigger benefit they drive to the to the business that's that's operating them. So that's that's really been kind of the key theme behind VM store is give you the ability to manage at that level instead of traditional storage where you're managing, you know, policies and things at a pool level which could have different types of applications or different types of workload profiles. Vm store takes totally different approach. We isolate it down to either the individual V disk within a virtual machine or a persistent volume. If you're talking containers. So very granular, very rich visibility and ultimately high performance on top of all of that. I know there's some people out there with Sans and possibly even vSAN still that are shuddering at just thinking about like, oh, I've got to manage this volume for 100 different VMs. And you're like, I've got I've got to try to set the right data protection. I've got to restore the whole thing as one big unit. I've got to set performance goals, you know, and they're trying to say, how do we how do we take our storage array kind of technologies and map it to this world of small, little increasingly fungible application components, right. And so, uh, let's talk a little bit about the world of containers and Kubernetes. Uh, you know, when we have a VM, you know, it makes sense to have a storage object that maps to that VM. Because I'd like to move the VM around. I'd like to move its data around. If I. If I run a data protect that thing, I want to protect the VM image and the data as a set, as a unit. And I want to think at that level. Let's talk about Kubernetes. Is that still the case? Um, so in Kubernetes it's a little bit different, um, in that you're working with persistent volumes as opposed to a attached virtual disk like you see in virtualization. So we've focused in on that individual persistent volume. Now because we're able to identify that individual object that gives us the ability to do things like we do for virtualization, like shows that we can do that at the individual persistent volume level as opposed to a mount point. When I say volume, I mean persistent volume. Um, so it gives you a lot more granular control. And one of the things that we're seeing, uh, you know, obviously we started out very heavy in the hypervisor space. And recently there's been a pretty big shakeup. Well, that's driven a lot of, uh, of enterprises to really look at different ways to deliver applications, which is now driving adoption of Kubernetes based deployments. One of the things that we're seeing is, uh, there's actually two things. One is visibility from the infrastructure operator's perspective. Kubernetes is new to a lot of these people. So when I say people, I mean the infrastructure operators. So they're getting new terminology and new new architectures and things thrown at them that they may not be familiar with. And one of the advantages of having Kubernetes and virtualization on top of a VM store is we provide a common language and a common ground for that new containerized admin interacting with the storage infrastructure. And now our storage guy can get in there and really see, hey, this persistent volume is attached to this container or this pod because we provide a browser of the Kubernetes architecture. So we're trying to add some tools for the newer Kubernetes admin to be able to be a lot more proficient using our tools in that space, um, than just traditional storage. The other aspect is visibility. What we're noticing is a lot of, uh, a lot of our competing vendors. Yeah, they can they can present a persistent volume. No problem. But there's no visibility as to what the performances of that, that volume, whether it's seeing latency. And then, you know, we get right back into the same performance trap that we saw with, with, uh, converting from a physical machine to a virtual machine. You know, we had physical machines, you could throw a ton of CPU because you already paid for it. It's in the box. But then when you go over to licensing that CPU, you now need to be more prescriptive in how you consume those resources. Well, traditional storage doesn't give you that visibility to see, you know, hey, I'm failing at the I o level. The latency is too high for this application. So you're kind of going troubleshooting really blind. Whereas with the VM store we give you that visibility. You can now see. Hey persistent volume number two that's attached to this application is experiencing high latency. And you can now go and determine, hey is that coming from the storage or is it coming from the network and be able to make a proactive change off of that information. So we're really trying to enable that newer Kubernetes admin, uh, or infrastructure admin that now needs to manage a Kubernetes environment and giving them a lot more rich tools to be able to do it. Uh, so I love these the, the utility of this. And you just brought up this important sort of angle to think about, which is not just matching the storage to the application component at a, at a very modular level, like a Lego brick level. Uh, but being able then to use the insight of how it's operating to troubleshoot, to look at performance, I assume, to capacity plan, uh, to possibly do some security work. Uh, and of course, I think for most of our audience, really, the key is to be able to protect and recover at that level as well. Right? So this is this is very, um. So does this, does this, does this allow people to do interesting things with Kubernetes that they weren't able to do before? Um, in terms of just moving services around, moving the data with the services, uh, and really in really becoming more flexible with it. So yes, we're laying some of the foundation for that. Um, so, you know, at our core, we have a capability to move data, we have a replication capability. And then Kubernetes has the ability obviously to deploy and, uh, and migrate applications. We're not necessarily migrate, but to, to redeploy. That's all based off of a deployment definition. So what we're working on in the back end, this is this is a little bit future, uh, casting here. Uh, but one of the things that we see is, hey, we can move all of the components and Kubernetes can redeploy it. It's all based off of a standard manifest that's already been deployed. Well, if we can capture that manifest and replicate it, we can then reassemble that application on the other side. So that's one of the things that we're working towards. It's not a feature we have today in the product. We do have the ability to snapshot and replicate persistent volumes, but right now we don't have the ability to look at a deployment as an example and say, move it from cluster A to cluster B. However, we are looking into that tooling and targeting that type of workflow because we do see the benefit. Now, one area that we can, uh, impact workflows today is around our our cloning capability. So if you have a workflow that generates a persistent volume, multiple persistent volumes, or maybe there's a standard gold persistent volume that is duplicated multiple times, that's something the VM store handles magnificently. So our file system is, uh, pointer based. So we can generate those persistent volumes extremely fast because it's we're basically creating a new pointer reference. Um, as well as it uses our standard replication. So if you're replicating one of these persistent volumes. Deduplication is at play, so we're doing it as efficiently as possible. So all of those existing VM store, uh, features and capabilities for the virtual machines are slowly coming over into the container space. So core items DOS, uh snapshot replication, which ties back into data protection. Those are there today, but you'll start seeing even more expanded capabilities. All right. Let's talk a little bit about QoS, because I mean I mean really my question is can I combine this with my what else I'm already doing with Tintri. If I already have some VMs, I already have some databases, I already have some other things that I'm doing. Can I just add support to that? Is it does it blend well? Yes, absolutely. So just like the VM store can handle multiple hypervisors simultaneously. Uh, we can also handle multiple platforms. So you mentioned it databases with Microsoft SQL and then containers with uh with Kubernetes. And quick little announcement here. We actually just released the CSI driver for OpenShift. So we can also do container support inside of OpenShift platforms as well. Um, but yes, uh, these guys can all go these workloads can all coexist together on the same platform at the same time. Uh, they can benefit in that they're on the same platform so that you don't have to send IO across the data center in order to complete a task, uh, as well as, you know, as you're going through refactoring applications, maybe today it's a three tier app in, in, uh, virtual machines. Well, you can take that first front end as an example. Uh, maybe it's nginx web server start containerizing that while it's still talking to the virtual machine database on the back end, and then you just iterate through that process until you've converted the application to be fully cloud native again, all coexisting on top of the VM store and able to leverage VM store capabilities. I mean, that's really kind of it's really kind of a powerful sort of feature that you might not think about at first like, no, I can do all this at once, right? And now I've got a lot of flexibility and ability to move forward and evolve. Um, so how does this scale is this scalable? When we talk about enterprises and reach. Yeah. So the VM store has had a, uh, traditional scaling architecture in that we have a control plane that is Tintri global center that's deployed, uh, independently of the data planes, which are the physical VM store appliances. Uh, a single Tintri global center can manage up to 64 data planes, uh, so 64 physical VM store appliances, or even some of the containerized instances we have with Tintri Cloud Engine as an example. Those will all fall under the same management. Now we don't scale as a, uh, a a large pool of storage. We scale as a federated pool of storage. So they are independent pools per storage appliance. However, our intelligence sits over top of it and is watching the monitoring or the statistics coming out of these boxes, and we actually are able to send back recommendations to the operator. Hey, this is a very write heavy workload and you've got a pretty write dormant VM store over here. Move this workload over to here. It'll balance out your overall performance. So we're able to give those performance recommendations back as we're analyzing the whole fleet of VM stores and the workload that's being generated over top of them. And this data is all being this this decision data is all driven off of the actual workloads IO, not some simulated model or an example. It's real world data that's being generated off the box to be able to make those decisions. Right. So let me just let me just ask you this to step back a little bit. Not everybody who in our audience is is a VM admin, but they all have sort of this IT perspective. What are some of the things that they could be looking for in their environment that would say, hey, maybe we should be checking out Tintri for this. What are some of the some of the some of the, I guess, bad smells, I guess you might say. Yeah, kind of the areas that I see that the VM store shining is, is in those, those multi use um, environments, especially where you're doing a concentrated application conversions. Um, you know there's there's other areas that the VM store has traditionally been been very good in uh anywhere where you can isolate I o paths and where I'm going there is around database architecture. So we do have SQL integrated storage, very unique way of handling that. But the way the box handles the I o architecture, you can actually build virtual machines that take similar advantage of that architecture. What I mean by that, remember back in the old days, uh, DBA would build out a SQL server, you would have multiple Raid cards in there with multiple disk sets behind them because they're splitting that I o path to each either database file, log file, so on and so forth. You can emulate that architecture over top of the VM store. Doesn't necessarily have to be a database. That's just a good example. But when you place an individual database on a v-disc, an independent v-disc that gets its own isolated I o path. So there's a huge performance benefit. So anywhere where they, uh, prospect needs to be able to see deep I o information specifically around latency and I o performance. Vm store definitely shines there anywhere where you're trying to create a consolidated infrastructure that needs to run multiple platforms side by side. Again, another shining use case for the VM store. All right. I mean, that's all really cool stuff. And I remember looking at, you know, the analytics for Tintri years and years ago where you guys are that old. Not to imply that we're eight year old. Uh, but it was very cutting edge, cool stuff at the time because as you mentioned, you can get statistics and analytics down to the application object. Yeah. Analytics has been, uh, a really awesome tool. Uh, we've got some capabilities within that tool around what if modeling. Um, so, for So example, you know, you need to bring in a new a new fleet of SQL servers to support support your next generation AI app. We can actually model that workload. What VM stores do you have today versus what are you going to need to be able to support that? Or if you're transitioning workloads, maybe you know that that project wasn't what you guys thought it was going to be, and you want to shift those resources to be something else. You can get in this tool and model and change around workloads and put different, uh, model VM stores in place and really understand the impact of which model or which configuration best benefits your application or your deployment requirements. I mean that what if capability is just really the heart of everything we want to be able to do? Being proactive, being predictive, uh, being able to optimize things, uh, and apply other forms of intelligence, you know, coming out to be able to, to have the right information first. Um, so, so there's a lot, a lot of cool things you guys are still doing. And, uh, and now that you're bringing along Kubernetes and the CSI drivers and stuff, this is this is just continuing to get better and better. Uh, if someone wants to find out more about, uh, tintri, what would you have them look at first? On maybe on the website or what kinds of research should they do? Yeah. So tintri. Com we, uh, we try to keep that as up to date as possible. Um, just a quick little outside plug. Uh, we've released a new 7290 model. Uh, that is the latest, greatest on our website. So please go over and check that out. Um, but all of our socials. So, uh, at tintri, depending on what platform it is, we, uh, we have a presence on all of them. And then another thing I could plug is, uh, we run a webinar, uh, usually about once a month. We call it a geek out, and we try to focus in on a specific topic that matters to businesses and then really jump in and show how the how the product can adapt to the challenge or the or the need. So really good resource there. Um, and then again, we've got reps throughout the United States. So if you uh, if you already have contact with a rep, please, uh, you know, get engaged with them, happy to jump on and do a call and kind of go through some of the product and get you up to date. Um, if not just, uh, reach into our general inquiry and we can get you hooked up with somebody as well. All right. Thank you so much. I know there's a lot more to dive into. Uh, that we just really scratched the surface of. I would love, you know, just even going back and talking about shows again for half an hour, you know, just like in the auto shows. Because I always I think that's really nifty to just go back and date myself. That was nifty stuff. Uh, but we got to go for today. But so thank you for being here. Thank you very much. It was a it was a pleasure. And, you know, always great to have a conversation with you guys. All right. And check out Tintri, especially if you are still struggling with storage for your hybrid environments, for your mixed environments, if you're still doing things the hard way with Sans and take care of guys.
