iXsystems’ Morgan Littlewood Explores Shifting Paradigms in Storage

This is Textron tv. Hey guys, thanks for the throw. We're here with Morgan Littlewood, who is Senior Vice President of product management and business development for IX Systems, and we're talking about open source and storage and why we haven't seen a whole lot on that front just yet, but maybe that's all changing with the rise of container storage.

Morgan, welcome to show. Good morning. Welcome to, thanks for, uh, having us.

We have seen open source be used almost everywhere, but it's very little of, it seems to find its way into the storage environments. Give us some perspective as to why that is and why that might be changing. Um, storage has always been a more conservative part of the IT stack.

Um, you know, data is critical. People are always looking to make sure that, uh, their data is, uh, reliably, uh, um, stored and, uh, you know, performance on access. Um, so in general, I think that people haven't wanted to skimp on, uh, their investments there.

Um, with, uh, open source, there has been some significant progress over the last decade. Um, uh, in our case, sun sort of started the process play out and sourcing ZFS and ZFS is becoming, um, or has become, you know, the most reliable file system out there. And that's been the basis of our work is to, uh, commercialize that, um, software and that capability.

ZFS make it available as open source, um, in the form of Trina, but also make it available in the form of appliance that, uh, enterprises can use. Um, which include all the support features that people need to be able to reliably operate their storage environments. Is the cost of storage starting to get prohibitive?

And that's part of the reason there's more interest in open source these days, because every time I turn around, it seems like there's more data than ever that we're trying to store. Yep. There's always data growth, that's for sure.

And, uh, unfortunately for most of the, uh, enterprises out there, their budgets for the storage don't grow. So they've gotta store more data, they've gotta store it more economically. Um, yeah, typically what we, uh, we find is that the raw storage costs, which are the costs of the discs and the flash drives, et cetera, have generally been coming down not so much right now.

Uh, right now there's a bit of an upswing in terms of prices of the raw storage, um, but the costs of the storage systems and the software associated with running storage, uh, haven't been going down as quickly. And so the, uh, benefits open source is you can get a free version or you can get a reasonably cost costed version or priced version with support. And that's effectively the business model we have is offer both of those and, uh, uh, help to grow the business that way.

And, uh, help to solve problems with, uh, storage budgets. Early on with Kubernetes folks were saying only run stateless applications and store the data somewhere else. It seems like we're seeing a lot more stateful applications lately, a lot more databases running on the platform.

What's driving that transition? Uh, well, I think it's just the, um, um, uh, effectively the, uh, use of Kubernetes and more important applications. Um, most applications that are important have stayed in some way.

Um, and so there are tools within Kubernetes to use external storage. So the CSI drivers, uh, let you use most external storage systems. And so a lot of, um, Kubernetes users use, for example, Trina as an external storage system for their clusters.

Um, and now what we're offering within, uh, Trina is the ability to run Kubernetes apps built into the storage as well. And so that's a, uh, uh, a change in terms of the opportunities, uh, available. Some people have a perception that the entire platform of Kubernetes is difficult to manage storage being no exception.

Is that getting any easier? And does ZFS make a difference in that regard? Um, Kubernetes has challenges and so clearly, uh, there's, uh, um, uh, some well reported, uh, uh, complexity there.

Um, uh, what's we tend to focus on is where people wanna have only a limited number of apps running on a system, then we'll make it easy for them to deploy those apps on a single storage system, uh, where it's a full site scale cluster, um, then it's the same as normal Kubernetes. We still have to, uh, maintain that Kubernetes cluster and then access storage, which, uh, we make easier to do, but we don't eliminate the complexity of Kubernetes, You know, on premises environments. Of course, we've seen storage administrators for years.

But with Kubernetes, is that becoming, uh, a job function that's part of the whole stack, it's part of the compute. And, um, I guess I'm asking who's in charge of managing storage in these environments? Uh, good question.

Uh, we tend to find that it is a mix. It's either the, uh, DevOps people running the Kubernetes stack and they're responsible for storage as well. Or sometimes it is a storage administrator that is separate from the, uh, other people running the communities cluster.

Um, I would say probably 50 50 at this stage. With your approach, can I use ZFS both on premise or, and in the cloud? Is the environment becoming more federated or just how do you think this is all gonna evolve?

Yeah, so ZFS can be used, um, underneath a, um, um, a VM or a cloud instance. So it can be used in the cloud, but most of the work that we do, and most of the deployments are on prem. And, uh, effectively it provides, uh, cloud like, uh, um, uh, reliability of storage, um, on We the rise of data sovereignty issues and data privacy, or are we starting to see something of a shift back to on-prem?

Uh, I wouldn't say, um, there's been a major shift, but what we do see is, um, effectively, uh, um, a couple of trends is that generally, yeah, early adopters, um, or people building new systems will often start in the cloud. Um, but once they have a large amount of data, the costs of cloud storage are actually quite high. And so what we do see is companies, as they mature, often bring their larger data sets, um, back into the on-prem, uh, either a colo or their own data center.

Are latency issues different with containers than they might be in a traditional monolithic virtual machine kind of environment? Is that, uh, a little more nuanced that people have to think through, Uh, for storage? Um, I don't think they're that much different.

Um, I think both the M'S and containers want low latency storage, particularly if it's a database type application. Um, so we see them as pretty similar. Uh, the advantage of containers is that it's lighter weight.

You can spin them up more easily and you can have more of them on a system. So, but otherwise, um, uh, they have very similar storage characteristics. Do containers want their own storage systems or can they share storage systems with VMs and for that matter, I don't know, serverless computing frameworks and whatever else is out there?

Yeah, They can share storage systems with, uh, other applications. Um, do you tend to find that people build dedicated storage for Kubernetes environment? Yes, but it is not required.

You can certainly run VMs and, uh, standard databases as well as container workloads on the same storage at the same time. Of course, these days cannot walk down the street without somebody leaping out to tell you about their great new AI thing. Will AI get applied to ZFS and storage in general, and what might we expect?

Yeah, um, AI requires some level of storage for various functions, uh, whether it's archiving of the data that, uh, is required or being able to provide a larger data set for, uh, some of those applications. So we get used in those type of environments where people are using, uh, um, doing AI for research and then using us for the data storage. Um, in the case of Kubernetes, um, some, uh, lower level AI workloads get put into committees type environments.

And so, uh, for example, in within Trina, you can share GPUs across many different workloads, and those GPU workloads can be, uh, uh, can include AI type workloads. As we kind of look into 2024, what should we expect from storage and containers and Kubernetes? You know, as you kind of get your little crystal ball out there, what are looking for?

I can't say I predicted everything that's, uh, already happened. So, uh, um, I do think that, uh, Kubernetes continues to grow. Um, particularly software developers prefer to work in a Kubernetes environment, and so we prefer to deploy that way.

So we are finding more and more apps continuously, um, being, uh, developed for a communities environment. So there is gonna be growth there. Um, the IO world, who knows, um, but, uh, it's right now mostly constrained by lack of supply of hardware.

Um, so the, uh, um, people either have to go to the cloud or wait quite a while for their, their GP infrastructure. Um, in the case of storage, um, we, we generally take the approach of, um, we need to be flexible. Um, the way that, uh, customers use their storage, it's changing all the times.

So it used to be bare metal systems accessing storage, then it became VMs, it's now becoming containers. Um, and in many cases it can be a mix and it might change, um, every year. So flexibility is the key thing.

We make sure we can support all of the different storage protocols, whether it's ice, Fs, SB or S3, and, um, um, uh, adapts the storage to meet whatever the application requirement is as it changes. Do we need more contributors to open source storage projects? It seems to me everybody kind of focuses on, you know, big projects like whether it's Kubernetes or Linux or whatever it may be, but are we overlooking storage?

Uh, yeah. So in the, the storage side, uh, open ZFS is the, uh, um, major, um, uh, group that we work with, and that has a very large number of contributors that work together. Uh, in the storage world.

It's, um, uh, quality is very important. So we need as many tests as we do developers, so, uh, if not more. So really the focus is on trying to grow a community that can test everything properly and make sure that we don't have no issues, uh, once the, uh, software is in production.

So what's your best advice to folks who are running these container storage environments? What's that one thing you see people doing that makes you shake your head and go, folks, we need to be smarter than that. Um, I would say that, um, the, the container, um, description is sort of a little bit, um, uh, too vague, really.

It depends on what you're trying to do with containers and, um, do you need persistent data that is retained for years or are you looking at just, um, uh, an application that, uh, needs persistent data that is retained for hours while it's, uh, doing a particular function. So, um, generally we're there to help for applications that need persistent data that is gonna be retained for years, needs to be snapshoted and managed and replicated and protected. And so if you need that type of environment, then um, Trina is a very, very helpful tool.

If you're looking more for a very compute, intensive, um, environment, um, then the matter, the storage you need is much less. And, uh, you can, uh, uh, focus, uh, on a, uh, on a more compute focused, uh, uh, environment. All right, folks, you're heard in here.

Not all persistent data is the same. Use cases matter, and you need to know what you're doing when you look at it in the first place. Hey, Morgan, thanks for being on the show.

Radio. Great to meet you. Bye.

And back to you guys in the.