Anjali Arora on DevOps Teams Embracing Platform Engineering

This is Techron tv. Hey guys, thanks for the throw. We're here with Angel Aurora, who is Executive Vice President for product and development at Perforce, and we're talking about how to make DevOps infrastructure more efficient because well, everybody's trying to save a nickel these days.

Anjuli, welcome to show. Thank you. Mike.

I feel could be wrong, but most of the developers when they're building software seem to think that the underlying infrastructure that they're using is somehow magically free. And the reality couldn't be anything, but, uh, organizations spend a lot of money on the infrastructure, on the cloud computing, the CICD platforms and everything that goes in between. Um, we have so many different DevOps teams in these organizations.

Are we just fundamentally inefficient? Sometimes it feels like, you know, the development teams are the IT equivalent of drunken sailors, but, um, you know, your sense of what's going on here and why are we starting to pay more attention? So I think it goes back into history of infrastructure and how it was done, how it is done today, and how it'll be done tomorrow or, and here tomorrow is like two to two to three years from now.

Okay? So how infrastructure was managed in the past was by it, but they became a bottleneck for development teams who were doing SaaS, and they needed to move faster as cloud came into being, and applications needed to be deployed in cloud as they were developed. So the DevOps teams came into being, and they started basically provisioning, provisioning the infrastructure, and they became part of the agile teams.

Now, that created the anarchy that you are talking about, the development teams thing. Infrastructure is free. Cloud is like mushroom, right?

It's, it's like a forest of mushroom for the, for the containers. Keep growing. The VMs keep growing in the cost in the cloud, which means your cost keeps growing.

And that's where platform engineering is coming back, which is really centralizing your provisioning of infrastructure through a set of policies, right? Which is infrastructure escort, security as a policy and your cost controls where the development teams now can do a self-service, but they're governed by a set of rules and policies. And that is kind of the current reality that organizations are trying to get to, but most organizations are not there.

So that is today's reality. So what happens in future is going to become even more interesting, but that's where we are today. So speaking of the future, we hear all this conversation about platform engineering.

Where does that fit in this conversation? Is one essentially a response to try to wrap our arms around the other, or is one, is the infrastructure just one piece of a larger conversation? Well, infrastructure is one piece of a larger conversation, right?

So when you think about IT, infrastructure is, to me, when we think about infrastructure, it's like the plumbing, right? That all software runs on. You need infrastructure, whether it is your on-prem infrastructure or your cloud infrastructure.

Um, and we can keep, we can keep adding the infrastructure, but in the end, we have to make efficient use of it. And what you will see happening is that these DevOps teams will actually become very reliant on the platform engineering teams and security and ai and all the tools will make it so that there's a closed loop system where all of the production data, the user behavior, all of the historical knowledge that we have will actually be used in development production and in making sure, um, that the infrastructure provisioning becomes adaptive. Uh, that anything that we are fixing, whether it is, you know, an outage or um, performance bottleneck or a security incident, it'll all start happening in a closed loop system, which doesn't happen today.

To your earlier point, we kind of invented DevOps to get out from underneath the thumb of centralized it. And so when we started talking about, uh, optimization of DevOps infrastructure and platform engineering, a lot of people read that for code of revenge of centralized it. So how do we kinda get DevOps teams that prize their independence to kind of buy into all of this?

So you will find in most organizations, the platform engineering teams are actually the DevOps teams, which are taking over and coming up with the best, um, set of practices and policies that they are implementing. And these platform engineering teams are more inside what I would call engineering or r and d instead of being inside it. So it may still own the infrastructure, but the platform engineering team is usually not in IT anymore.

So, so control has gone the teams which are developing the applications. So they are recognizing the issue and trying to solve it amongst themselves in a way that requires some negotiation of standardization, right? Ultimately, I have to come up with some sort of, uh, agreement as to what we're all gonna use.

So how does that conversation proceed in your experience? So I, I can tell you some kind of experience based on like what we see even in our organization, right? So we have it, we have r and d as an example, right?

And in the past, um, the development teams had to depend on IT to provide the infrastructure, and that slowed everything down. Now we have platform engineering, which owns the platform, which the developers, uh, and the agile teams actually self-serve their platform form. And the platform engineering team is actually negotiating with IT, with security, which is our CSOs organization.

And with development in making sure they know what is needed by each of them, because the IT teams want to make sure we are making right use of the infrastructure, right? They want to control the cost, the security teams want to make sure the infrastructure is secure. There are no, no, you know, gaps, um, in the policies.

And the dev teams want to move as fast as possible. So the DevOps, um, engineers who are part of these agile teams work with the platform engineering team, um, to self serve for the developers. I, I hope I answered your question.

So, so there is a very, so think about it as there's a very close collaboration between the IT team, the CSOs office, which is a security team, right? Which is setting your governance and policies and the r and d teams or the development teams, which are actually creating the applications and they're working with the DevOps teams to deploy them. How much do you think that the rise of AI applications is forcing this conversation?

'cause it seems to me the amount of infrastructure we need to consume to build the next generation of software is, well, about several order two, several orders of magnitude larger than what we had before. Uh, it is, um, I remember a time when it was you worked on one computer and now Yeah. And that one computer was hard, right?

My phone has more power than, than anything that I had, you know, 30 years back. So anyway, uh, data fight, um, AI is, uh, actually, I don't think we've seen the impact of AI yet. AI will actually force this collaboration to increase, and AI will not only force the collaboration to increase between these teams and agile teams.

AI will become a part of the conversation. So AI will become, to me, AI will become a team member and it'll supercharge every team. And AI will take over anything that is repetitive, you know, like you need to repetitive or that requires a lot of grunt work.

Like, you know, analyzing a log or, you know, creating a test or, you know, doing anomaly detection of in, you know, your, from your logs of a security issue, AI will do all of that for you, and it'll work with all these teams to give them options to move faster. So in an odd way, AI creates the problem, but AI will also help solve the problem. AI will definitely help solve a lot of problems where the teams will focus on really continuous improvement.

They will focus on kind of evolving their architectures. They will, they can focus on innovation and uh, and AI will handle things that it is good at doing, you know, finding kind of, um, patterns, predicting, giving, fixing things, you know, giving options to people like, uh, let's say they have a, um, what is it call a runbook, right? Giving them options on these are the different things, uh, ways these runbooks you can run to remediate this and this is the best possible option.

And, you know, then your res can say, okay, run this one for me. So from your perspective, are we on the cusp of some massive rationalization of the DevOps tools and platforms, or are we gonna continue just to have our choice? We're just gonna be smarter how they consume infrastructure DevOps tools of today, if they do not embed AI inside them, and I'm not talking about, um, AI like chat g pity, right?

Where you are asking questions and you're getting help. I'm talking about if they do not have AI agents inside them, which are removing the repetitive tasks, providing insights, and helping the humans who are working with them to elevate their performance, those DevOps tools will become redundant. Those, so the DevOps tools have to become very, very intelligent and utilize AI to do things like, you know, predictive insights to, you know, adaptive infrastructure provisioning to like ai, augmented security.

And I, I believe that's what DevOps tools will do. And they will be DevSecOps tool. They will not just be DevOps tools.

So what do you anticipate will be the biggest challenge to accomplishing all this? 'cause um, paper, at least it all sounds perfectly reasonable, but people are people. So what are we gonna do to make it happen?

So the challenge we will have is the, um, mentality of the people, the culture of these teams, and the appetite of the organizations to adopt ai. You know, there is a lot of, um, there's a lot of, um, what's the word I'm looking for? People are scared of ai, right?

They think it'll take away their jobs. They're also scared about hallucinations of ai, ai, they're also scared about, you know, what kind of data is AI using? So we will have to get people comfortable with that.

Um, we will have to make sure, you know, the organizations have a culture of accepting AI as a team member. Governance will be very, very important. Anything done with AI have to have very high level of governance.

I think humans need to understand, and I'm talking about any human being, and this is not just in DevOps. This is in, in any area, if they use AI as a companion, as a trusted partner, they can elevate their performance and any team which uses AI can elevate, they, they basically supercharge themselves. And I think that's a cultural shift organizations will need to make.

But again, governance, the, the, the risk tolerance, making sure what they are using is safe. Um, that I think organizations which are embedding AI will have to make sure they can, um, convince their consumers of that. So do you think that we will assume be deploying a lot more software than we used to faster than ever?

Because there's a lot of people saying, you know, with AI right now we're writing a whole lot more code, but it's not really making it out the backend necessarily in terms of more applications. So when do we see the benefits? I, I'm going to, I'm going to be careful about what I say here.

Writing a lot of code is not the same as having useful applications that solve a problem. People want solved, right? So I think we will see a lot more, um, business relevant applications come out that will solve big problems.

And AI will not only help in creating them, it'll also help in providing better outcomes. I do think we are a few years away from where the velocity of that will increase, uh, because there is a lot of, there's a lot of experimentation going on right now, and people are just generating things for generated to just say that we are using ai, but to, to build something useful, takes a lot of thought, takes a lot of research, and, and, and, and you have to validate it with customers. Mm-hmm.

So let's bring this full circle. Do you think that once we start rolling those applications out, that there'll be more valuations of how efficiently the code is actually consuming the infrastructure? And we will measure developers on that much more so than we have in the last, I don't know, three decades.

I've thought a lot about efficiency of other functions. Like, you know, when you think about there's developers, there are testers, there are, um, DevOps, SRE I've given it a lot of thought on testers, DevOps and SRE. I think those are the ones where we will have the first impact of efficiency and what they do.

My gut says developers will be the last ones, uh, just because it's kind of, you do a shift left and you know they're on the left most, uh, but eventually it's not only going to be, uh, the efficiency of developers, actually there's going to be efficiency of product people as well. You know, product managers, designers who are actually designing these applications or coming up with requirements. Everything will be measured.

Uh, and you will be able to say how efficient we were, were we from requirements to delivery, which is really hard to do today, Folks. No one's quite entirely sure how well this is gonna play out, but there's one thing for certain change is coming, and at the end of the day, somebody's gonna start measuring the impact of that change. So be prepared.

Actually. Thanks for being on the show. Thank you, Mike.

And back to you guys in the studio.