 This is your host Abhinav Harkya and welcome to another episode of our let's talk and today We have with us once again Michael Kade global field CTO of cast and by beam Michael is good to have you back on the show I'm happy to be here and today. We are going to talk about AI workloads in Kubernetes and their impact especially in 2023 before we Go deeper into this topic talk a bit about where do you see? Kubernetes in terms of maturity and running in production as an ever-growing scale every time I've been on there I think we've been we've been talking about those percentages and the gains that we're seeing at least from a Especially from a customer basis where obviously we're focused on data protection and data management within Kubernetes We're seeing a nice increase in in the footprint of customers that are moving or adopting Kubernetes as a as a platform now let's talk about AI based workloads Running on Kubernetes or in other way because when we look at AI ML It it works both it can all also be seen as a workload and it can also be seen as something that can actually improve a lot of technologies So so talk about it. How do you look at from what perspective? fundamentally AI ML whatever or deep learning as well like they all stem from data and data access now We're obviously going to be focused on Kubernetes and what our talk track is there, but this isn't new from like previous platforms such as VMWare or virtualization as well, right? So where we start we start talking about the data We start talking about the importance of that data without data You don't have an AI or you don't have a or you might have an algorithm But you don't have a data set to run the algorithm against how important is that data? Where do we get that data from and I think that's where that's where I tend to start Especially when we're talking to customers that are leveraging AI and they're building their own AI and ML operations or applications then we start talking about the data the importance of that data and Not always if you think about like a traditional environment where maybe you'll have a a database or a data service Maybe you'll have a some observability metrics and it's not as fun as maybe chat GBT or gaining insight into into data that way, but Observability has a a dream of different like informative data points So you might be looking at okay. How could I leverage that data? How could I get something from that and AI is sat on top gives us that insight into that data So instead of doing that from the production set, I want to take a clone of that data And I want to spin that up in the most economic like Efficient way so that I could then run run my algorithm against it or just basically get insight out of that data Without affecting anything in production. So I think that's that's kind of where I'd first start that process of Data's important and data is the fundamental part of AI and ML Ops if you look at Kubernetes once again The use cases are Beyond what the the creators thought of you know, it's it's just you know exploding folks are using it in so many different It's like initially it was it's state less than it is stateful. You're saving data there Which is also kind of though the Kubernetes community they have you know, they're maintaining it the way they want it But these new use cases they do affect project You know a lot of you know other sub projects and just in project also come up talk a bit about how is AI ML based workload affecting The Kubernetes self if it is yeah, I think I think the biggest thing is is that Me for one the personas that I'm speaking to aren't just data protection specialists. They're not just backup teams They're not just virtualization admins or sys admins. I'm now having conversations with data scientists How could we leverage that data? How could we move that data around and then obviously if I'm having them Conversations that and I'm right at the bottom of the stack I'm generally the the last port of call when it comes to data protection I'm like the least exciting thing on the on the docket when it comes to spend it spending on that on a new IT project But if it's getting to me then then people are like data scientists are also intrigued about Kubernetes which is only going to grow that community up into the use case that are available to it The challenges that that's going to bring is well if once this admin wants this and a data scientist wants this How do we? Make sure that we've got enough cycles or or maintainers to make sure that all of these projects have the capabilities involved Just going back to kubecon Detroit The big the big takeaway there was that we don't have enough maintainers or we don't at least look after our maintainers enough now And we have this flourishing community of growing into different Areas and industries we have The AIML ups around data scientists, but then you've got to think about like edge and iot You've got to think about the traditional virtualization admins, then the cloud engineers the dev ops engineers sre's platform engineer. It becomes a huge group or a huge community that are pushing for potentially different things or similar things But there's some differentiators in in each of those things that they're they're pushing as well. So I think I think the constraint is we don't have enough maintainers. We don't have enough people to Create and make what we need to to push all industries forward um But hopefully with what the kubecon and the the linux foundation are doing is that that's going to encourage more people to contribute and to maintain projects and I mean the the cncf landscape isn't getting smaller. I think I don't know the last time when you looked at that cncf landscape, but Yeah, there's still A large amount of different projects on there that are are constantly evolving To suit to suit the needs of all all members of the community. So first of all, you know, there's already, you know, kind of, you know shortage of When we look at data scientists and then if you look at Expertizing these communities and all those skillset. That's also not a skill set you can find easily so talk a bit about what does this economic social change that is happening will have on teams and We can also talk about, you know, what vm keston is kind of doing all the whole cloud native ecosystem is doing To help companies so that they can continue to move forward. Yeah, I think I think the skills Shortage obviously we've spoken over the last couple of years about the skill shortage around kubernetes and cloud native as a cloud native ecosystem is is apparent I think what the if we look at the world and the economics that's that's going on here is is very much Dictating that we have to do more with less which is always the case, right? When we're all when when interest rates rise and We have to do more with less and teams get shrunk and we have to do the same amount and and that's just that's the way this This world seems to seems to run And I've been saying about that I actually did a session at the beginning of the week around how the sysadmin is evolving that systems Administrator that maybe just looked after Probably shouldn't say just looked after a server farm within a data center And then progressively had to look after exchange sequel had to look after virtualization when that came along had to have that learning curve and then had to start looking at cloud now There's a lot of different platforms that were just mentioned there the sysadmin that I remember I was a sysadmin back in the day and I looked after a rack of servers I had never had the concept of virtualization. I never had the concept of cloud per se And I had to look after them and then you have to evolve as that that sysadmin then becomes a virtualization consultant and so on and so forth And I think that big next evolution is around automation. How do we how do we simplify that automation? How do we encourage people to use infrastructure as code? And it's very difficult because with virtualization you could instantly see that that That abstraction away like I can take that one physical 2u server or 3u server and now I can fit 10 20 30 virtual machines within that you can see that level of Like abstraction there and with the cloud you could see that you were sending workloads up and you were taking away a You were abstracting a layer of manageability of that that workload whether it be virtual machines or databases as a service You're abstracting a level of I don't have to worry about the operating system anymore I don't have to worry about the the server in the rack and the power to that I pay someone else to do that Now with automate automation there's no Real managed or abstraction layer around other than your own time As soon as you see For example, like using terraform from an infrastructure as code point of view And you create 10 virtual machines from just saying terraform apply That Shows that the abstraction there is your time. We're abstracting away that time now I understand there's forces of nature that stop us being proponents of change But by being a proponent of change and looking at infrastructure as code or configuration management around ansible around chef around kubernetes It provides us the ability to claw back some of that time because we're being asked to do so much more And whilst doing that you're ticking the box of being so much more valuable as a person Within the it industry or in the tech industry to be able to offer so much more out there as well So there's a massive challenge. There's the there's that I've been a I put all my time and effort into being a virtualization admin Um, I don't want to change again. This is never going to last. This is never going to happen And there's a lot of that. There's a lot of pushback People my age don't want to retrain and learn new new traits And I get that it is it's it's overwhelming. It's sometimes daunting It looks like you're just looking at ever is and I've used that terminology a few times around Everest looks like it. Well, I shouldn't say that I've never climbed Everest for the record But ever is from base camp and even even from the airport looks an incredible size and listening and speaking a lot to authors that have traversed Everest they actually say as you get to base camp and then you start going up It becomes more manageable, especially those people that have been able to to do it And I think you have to set that that path to that learning journey as well If you look at yeah, I am talking a bit about, you know, uh, what are the trends that you're seeing where you feel that Hey, what these are the some of the opportunities that we missed and uh, but now you see some positive trends when you will see that We will be leveraging a iml in a way that we should have So if we look at like technology over the last at least 20 years that I've been in tech It's always been driven by or mostly by that commercial sector as in A consumer. I'm a consumer of technology at home. I have my have my playstation 5 I have my nintendo switch. I consume technology. I have all my home automation Stuff running here and then we see that drip into the enterprise or vice versa But mostly it comes from the other way and I think The the the characteristic around like chat gbt that that's not blown up because of Like a couple of us in the it industry looking at it and seeing what it can do Looking at a massive data set from 2021 And seeing it all over my youtube or all over my like blogs and and and feeds that i'm watching But it's actually people that I see like down the pub and like people that are carpenters That I speak to on a daily basis friends. They're consuming that technology and that's the that's the uh, the Catapult for me is that as soon as you see People outside of the it industry Adopting technology we start to we will absolutely be double downing on that from a technology industry point of view I think chat gbt just shows the power of what that ai does. It also enables that automation level Like I don't know how much you've done with with chat gbt But you can quite easily make it It could it can write pretty good code samples if you ask it the right questions You can write a program with it. So that opens up the door to Again squashing that that barrier to entry that that learning curve that that i've mentioned before I think also Other big vendors bringing out their VR and their ability to to to jump into an immersive world Only opens up the door to bigger tech firms doing more of that things that stuff as well I think that is actually there I think we might talk about it like we've been talking about ai and ml for for a while about cleaning data Leveraging data and especially from backup Backup data we've been saying about doing this for for at least five years. I have at least but until something like a company like apple or microsoft or place or sony bring out a The capability of being able to do it in your own home That's what then leapfrogs us back into a oh, we'll start looking at leveraging data Getting some sort of insight out of that data to provide better business outcomes or Just like let's break it down to mitigating risk or reducing cost Like any of them three things is a top top thing for a business Especially in the day and age where we are where we're trying to reduce cost We have to do something with nothing We have to mitigate risk still because those threats around cyber security are not going away How can we be sharper to the the point around that and anything that can offer better business outcomes from that data? That allows us to understand a little bit more about selling you selling Trends and how people use our application where they use it Like they allow us to enhance that application Which means that we're only going to make more revenue in these tough times where people are People are like bracing for a for a huge economic downturn Talk a bit about the challenges that are there when it comes to as you rightly said it's all about data Challenges that are there and how casting them are helping you know users to kind of Protect some of the data because that is also important. There's two two Answers to that and one is around Just being able to protect data Just normal data services databases that are living within or outside of the kubernetes cluster as well as the whole Theme data platform looking after all sorts of data protection The second is obviously we focus on backup and recovery of that data But in doing so we also lend quite nicely to being able to clone a copy of that data From the backup versus having to use your production So if you are looking for that data set or potential several data services and you want to put that into one One application to understand a little bit more about that workload and what it looks like Then we've got the ability to lend that to it. So like I mentioned around I'm speaking to a lot more data scientists. They're actually using that sort of clone Technology to be able to push their algorithm against that and then glean information from that. So Again, we're just focused on the data and backing up that data from an insurance policy point of view But there's such an extension and extensibility to what we're doing that opens up the door to different areas and different people that have similar but different Requirements when it comes to that that data management How much awareness do you see is there when it comes to a ML workloads? And data production data recovery disaster recovery high availability So I would say since our last so we have a we do have a case study out there specifically around data scientists using casting k10 I would say that that that conversation is very much started even more so over the last nine months I know that The cube con in detroit was very themed around a i ml ops if if you will as well as data as well If we look at the release release cadence of kubernetes that three a year Each one of those at least for the last two maybe two and a half years has been focused around data and and data at scale as well obviously there's a lot of Scalability within kubernetes as a platform Compared to something like virtualization So yeah, I think again, it's it's a growing number that we're seeing around adopting and I think from a scientist's point of view They they don't know what they don't know and them trying to take models and and uh algorithms from maybe an existing Area that they used to work in and trying to bring that into kubernetes It's just that maybe a little bit different But also being able to get a hold of that data in a fast efficient way is is another key area to to consider Michael, thank you so much for taking time out today and talk about this topic today And as is all I would love to have you back on the show. Thank you. Thanks for having me