Meeting Software Development Goals with DevOps and MLOps – Satish Iyer, Dell Technologies

This is Techstrong tv. Hey guys, thanks for the throw. We're here with Satit Iris, who's vice president and general manager for emerging services at Dell Technologies.

We're talking about managed DevOps and managed ML ops and all these good things that a lot of people don't realize that Dell does. Atish, welcome to the show. Thank you.

Thank you, Mike. Happy to be here. So what is the genesis of this service that you guys provide around managed DevOps?

A lot of folks don't realize that you have this capability in the first place. Yeah, so, so thanks again for the opportunity, Mike. Look, uh, I think the capabilities, the two of them you mentioned, our primary goal is to develop services and take some of the pain away from a developer's standpoint.

Right? And I think, um, uh, our, you know, we actually within services, we do quite a bit of managed services. Um, but our goal primarily in this one is as, as how, how can we make some of our offers developer friendly, right?

And the two things you mentioned is, one is primarily to get developers to be doing more coding and worrying about less of infrastructure and other management. The other one is how do you make it more easier for them to actually, you know, develop, train, and do model management and so on and so forth, so that they don't need to worry about the underlying infrastructure required for the machine learning. So both of them, I would say in summary is to take some of that pain points from developers away and make their application development aspects easier.

Now, these services have their roots and capabilities that you use internally, so they're not just something you guys kinda stitch together and you've been using this for quite some time, right? Yeah, especially the, like the Dell managed developer cloud, um, is uh, is a collaboration between, uh, Dell services team, which is my team as well as also Dell Digital, which is our IT teams. So, uh, the Dell digital team has been quite, you know, obviously they are a proxy for a big enterprise IT right organization.

So they've been doing, uh, quite a bit of that, uh, with, with, you know, within certain, uh, they have the same indu, same problems as any company, any big IT companies do. So, um, they've been dealing with this problem. So the way we collaborated was to say, Hey, look, I mean you can be my internal customer and you tell, tell me what the problems we are to solve.

So it, it is a great collaboration between us and we use them as our proxy, but, uh, absolutely they're own internal customer and our first customer. Are folks looking to you to run the entire DevOps platform for them? Or are they looking for you as kind of like a bridge where eventually they help take it over themselves and they're looking to you to provide some training and expertise?

I, I would say it, it's, you know, it, it's not, let me look. It is a managed service and our goal is to drive more automated operations for the two basic things developers do today. Right?

Basically, how do they create workloads application developers? Either it's either a containerized ecosystem or they basically use virtual machines, right? The primary goal for us is to drive more productivity, allow more self-service, allow them to consume and manage more virtual machines and containers in a much streamlined API based on one that, that's, that's as simple as that, right?

And we would allow for, you know, Dell will manage the VMs and containers throughout the entire lifecycle. That's the question you're asking. Cuz I think to me, if we delve us that end to end that allows these, you know, their devs to actually focus more on, you know, what they think they need to do, which is basically coding, right?

And our goal is to simplify the, not only the operations, but also, you know, financial management and the control and the, over the entire infrastructure. So we would like to basically take all things which are, I would say, mundane away from them so that they can focus on what they need to do. Right.

Are you seeing a lot more interest in DevOps among your client base? Because a lot of what you guys historically have done is around VMware and virtual machines and there's a lot of IT administrators. So what's the, what's driving people on the interest in DevOps?

Uh, I think two, two fundamental things, if I had to look at the market broadly, right? Um, uh, the first one is, you know, we are truly, truly going into multi-cloud, right? Um, and I think there has been a lot of discussions around, uh, cloud in the last 12, 15 years.

Uh, you know, there is this massive shift to public and then, you know, we said, oh, customers have to have a hybrid cloud. You know, there is a lot of opportunities in color. I think in the recent, you know, if you look at what, you know, a lot of big enterprises are looking at, they are looking at cloud and they're looking at truly to say, okay, they are making the decisions on what applications, what workloads will run there.

Um, they're not, you know, rushing madly to one another or another, right? I think so the notion and the recognition that it's a, you know, it world is gonna be a multi-cloud environment, I would say it's a first major thing. And I know it's not sudden or gradual and sudden, but it's been quite gradual.

But I think we at the point where companies post covid are recognizing that. And then the second aspect is, you know, along the same lines, I mean, that basically means customers are optimizing, right? So customers are saying, you know, um, there is quite a bit of repatriation going on, especially to do with some of the cost aspects, big enterprise.

It haven't really moved a lot of the workloads into public cloud. Lot of the big applications are not really what you and I would call like very cloud native anyway. So I think, um, it's a mix of both.

Um, so when we, when we talk to these customers, we, you know, again, this is primarily for big, um, you know, top 5,000, 500 big customers, this, that, the complexity, and they have a lot of ecosystem within their own environments for developers. So I think for them to, you know, bring that cloud experience as on-prem is, seems to be paramount and probably not going all the way to a public cloud, but I think, um, they know what the ease and the flexibility is when they actually consume certain things from a hyperscale environment. Um, so I think we are trying to strive to provide that with an on-prem.

So I think it's the multitude of things, but I think that would be the, to me, the multi-cloud is probably just beginning of this, uh, interest and discussing. Do you think that the infrastructure environments, especially in a cloud native era, have just gotten too complex for a lot of organizations and it's easier to lean on somebody like you? I do think it's getting quite complex.

I think the, you know, um, uh, even if I go to any hyperscale environment, you know, there is so many opportunities and so many options for me to even choose a compute note and a storage note. So, um, you know, you can actually talk about what level and, you know, you know, you have not just memory and storage and compute optimized, but you have multi, multi variance of those. So I do think that there are maybe been too many options and sometimes I think it's, uh, uh, again, as, uh, one of the top infrastructure providers, especially in the data center space, right?

Um, I do think that we like to keep things a little simple. Um, and I think when another aspect is, look, I mean, customers have big environments today, right? So when you go tell them that you can provide that ease and the flexibility, which they're used to in, you know, let's say in the cloud world, uh, I think that's a very strong story and our customers like it, uh, because now, you know, they're, they're, they already have a multiple embedded eco, you know, hardware infrastructure ecosystems today, and it is easier for them to say, you know, gimme a layer of piece of software which will allow me to actually bring that to my current, my development minus, Is there also more emphasis on application development?

And are folks kind of shifting and realigning their IT budgets a little bit so that they can free up more dollars to build more software? I think so. Uh, I, I, again, I, we, I touched on the repack part a little bit, right?

I think, um, there is, there is quite, I would say conscious, uh, there is quite a bit of, uh, um, understanding on cost recently in the last few years than before, uh, especially in the last year, um, given all the things we are going through. So it, it absolutely financial management and, uh, visibility into cost, visibility into consumption, and whether they're utilizing all the efforts they got is actually top right. I mean, customers are thinking that, um, and they want that flexibility, right?

Uh, but if you also think about it as, uh, enterprise IT as a, you know, single source by which they will actually have various business units consume infrastructure, and they're also looking to say, you know, how do I make sure that multiple bu within my company and within my enterprises are consuming my own it? So I think that has been a conscious effort now to look at the bottom line. Yeah, Of course, you can't walk down the street these days without somebody telling you about their great new AI thing.

Do you think AI will be applied to DevOps and what might that look like going forward? I, well, first of all, great question. Um, because, you know, can ML ops, especially what we call, you know, DevOps for ML is not new, right?

Um, that existed for a long time, right? Um, and you know, you've been talking about it, writing about it in this space quite a bit, Mike. So I think to me, the notion of, um, managed ML ops, um, where you actually have some amount of, you know, tested validated platform so that to drive machine learning workloads, right?

Um, same concept, right? Take the pain away of managing the hardware and the software so that, you know, developers can reduce the time to value for their models. I think that is important, and it becomes more important now with all the, I would say, buzzer on gen AI and l LMS than others.

Because enterprises are looking to say, what is their out in public? What can, what data sets do I have internally? How do I run some of these models which are applicable to my own data sets internally when they do that, or when they start thinking that, I think more enterprises are thinking that now than before, that makes it important for them to have, you know, an environment where they can actually do DevOps for these mls.

And you know that this is not easy. Most of the models, I would probably say less than 20% of the models see the light of the day. Most of the models which developers develop are not even deployed.

So I think it makes it, and that's because they're spending time more on, you know, worrying about the, the, you know, management part of the platform and the workflows than actually building models. So I think that's another area where we should definitely think we have an opportunity to coplay and make, uh, the developers life easier. You think ultimately DevOps and ML ops may converge because it seems like they are enjoying at the hip and they both have at least a set of best practices.

I mean, one's a little more appealing to a data scientist mindset, but ultimately these two teams need to work together, right? I absolutely think so. I mean, I, I think it has already happened, right?

I mean, you know, ML ops is the way for tool DevOps for very specific environments for learning. Um, I think, you know, the notional, um, the thought about reducing manual operational tasks and things we have to do, um, and you know, and with a very streamlined way is there, right? And, you know, and going back to the earlier conversation, I mean, customers are used to doing things in, you know, if they actually go get an ML platform in a hyperscaler today, right?

So you can go to any other major hyperscalers, and everybody has their own really good ML platform. So a lot of our customers are used to that. But now when they're saying, how do I take my own data sets internally and apply some of this, and I don't want to go take, push this data out to any public cloud, then, you know, that becomes important.

So they have experienced what good looks like, and I think now the question is how do you, how do you kind of bring it back home, right? I think that's, uh, you know, how do you bring the enhanced security and compliance and a lot of that stuff, which I would say they don't think too much about in pub when they go to public cloud now becomes important for them to think about. So what is that one thing that you see when you engage first with a client that kind of just makes you shake your head a little bit and go, geez, I thought we were better than this folks.

I mean, you know, what's that one thing you see folks doing that you would give them advice to say, Hey, get this together today, and you'll be in a better position for DevOps tomorrow? I, I, I, you know, I don't know if, uh, I can say that as an advice to a customer or a client. And as you know, you know, customers are always right, but I, I do think that the important thing which we need to think about is, um, this is not, especially on ML and ai, the technology problem is, is there, but it's not a technology problem alone, right?

So it's a, there is a lot of people process problem behind it as, you know, as somebody who's looking at it from a services lens. Um, you know, not just as a somebody who builds a piece of software or a product and gets it out of the door. I think it's important.

There's a lot of things which has happen behind us. So I think it's important and customers are recognizing that, right? So we know a lot of the customers have been doing things a certain way for a long time.

So just because we bring in a gen AI or an L l M model to do some cool things, it'll take them some time to adopt that within their enterprise. How do they change that? How do they take that and drive business processes downstairs?

How do they actually change the way they do business based on whatever the AI outcomes are? Those things are not simple, right? So I think, so that's one area I see some blind spots because, you know, everybody wants to jump in the pool, like you said.

Uh, what people, you know, for enterprises are slowly starting to understand is, you know, it, it's just not a tech problem, right? So it, it's great you have these tools and you, it's great. You can actually have some of these decisions which are much more faster, smarter, and more cost effective at some point.

But it's also that how do you actually take those and apply to your underlying business processes? I think that's gonna take some time. And I think that's kind of where, uh, I think the realization will be the next six months, six months later here.

All right, folks. Well, you heard it here. When it comes to DevOps, you don't necessarily have to go it alone.

Satish, thanks for being on the show. Thank you. Thank you, Mike.

Happy to.