Vishal Raina & Shibashis Mishra from Adobe on AI, IDP, and Zero-Downtime Migrations | swampUP 2025

Hey everyone, we're back here. Hey, this is our last interview for Swamp Up 2025. I think I am gonna wear my Jfr hat.

How's that? Oh, You look Beautiful. All right.

I won't mess my hair up. Yeah. Um, so let me introduce you to our guest for our last interview today.

There are two folks from Adobe. On my far left we have Shiba Shiba, she Msra. Thank you.

We call 'em Shibu. Shibu. Yeah.

I go by Shibu. Yep. Shibu is here.

And to my immediate left, we have Vishal Reyna. Yes, sir. Right.

And we're just gonna call you Vishal. Yes, sir. No shortened names there, gentlemen.

Welcome. And thank you for being our last guest here on Swamp Up 2025. Before we go further, I mentioned you both with Adobe.

Mm-hmm. And we, and as we were talking offline, we actually did a whole, what I considered a great series of interview with the Adobe security team around security and ai and how Adobe kind of eats their own dog food, if you will, or drinks their own champagne around security and to secure the products because all of us have Adobe accounts and we don't want that information getting Hack. Um, and we did a series of articles on it, you know, it was a whole treatment.

But you guys, you know those fellas from the security team? Yes. But you're on a different team.

Correct. Tell us about your team. So I'll introduce our team.

Um, we are the platforming group at Adobe. So when any developer wants to build something and ship it to their customers in one way or the other, we are providing the capability to those developers to make it happen. And, um, we, the pa the organization is named developer platforms, uh, and we partner very closely with Security organization, uh, led by Brian and Sure.

Chlorine and others who joined you in the past. And, uh, we work with them closely and, um, yeah, that's, that's our mandate, help our developer ship software faster to our customers, better software. So when you say developer platform, is it like a true IDP At this?

Yes. At this. com.

Mm-hmm. org group. Awesome.

Um, so I know a little bit about Enough to get in trouble about IDP. Is the IDP sort of like based on like a backstage type of Yes. Thing And is it, do you want to talk about that or, I know it's not really part of what we were gonna talk here, but I'm curious.

I can talk about it. I think, um, in fact, Adobe was one of the first companies who worked with the industry, and we actually put out a lot of content back in 2022 when IDP as a term was picking up. And yeah, we go and look at, uh, the different journals, I'm sure including some of yours.

You would find that Adobe had made contributions. So we were the first few adopters of the IDP concept, and we implemented our company. We built it on top of open source software like, uh, backstage, which you talked about.

That is the portal that our developers use. But it is powered by a lot of the CNCF community open source software. Um, we are a big consumer of, uh, Argo.

Yes. Uh, we are a big consumer of Kubernetes. And then many More, well, the, all the backstage stuff has Kubernetes at the heart.

Mm-hmm. And then of course the Argo GI Ops. Yeah.

And all of this is now, you know. Yeah. It's, so I'm proud to say that we are probably one of the biggest, uh, installs of, uh, GitHub software in the industry.

I think we really set really big scale. We have really collaborated, uh, very closely with the community to scale the infrastructure. In fact, there is A-C-N-C-F blog that we have published on how we have scaled, uh, Argo setup for ourselves to meet the needs where the, the off the shelf, the open source software doesn't meet our needs.

Uh, and I'll probably send you the link and you feel free to forward Please do. Well, may not be what we wanted to talk about. Hey.

But Q Con is coming Absolutely right. Q Con cloud native con will be, uh, I guess in November. Yes.

In Atlanta. And of course we'll be there live the whole time. Um, so, and that'll be both for our cloud native now and platform engineering sites.

So maybe we'll, we'll talk more about that. I need to shift though 'cause we are here at, But J Rog Jfr and talk about what you're doing at Swamp Up. I'll let that, uh, Vishal.

So, uh, We, we did a database migration. Um, and as you know, most of the database migrations need downtime. We were able to accomplish a zero downtime for our end users, and we wanted to come in here and share the best practices with our peers.

I thought because we don't have AI in our subject or presentation, we are not going to get any audience. Turned out that room was full and we got a lot of exci engagement right up to the point where we were meeting co couple of industry colleagues here. The goal was like how, how we came up with an eight hour downtime and that got shot down and how did we pivot in less than 28 days and made a zero downtime using cloud architecture, um, to accomplish a z.

And many of historical Jfr Artifactory users who are on MySQL or Microsoft, um, SQO will need to do this to come to the Postgres. And we shared our best practices, both in terms of how we solve it with technology and also best practices in terms of processes and people and communication. And at the end how we accomplish that whole, uh, zero downtime.

That's A great case study, a great case study. Let me ask a question just between us. Yeah.

Um, is the fact that it didn't have AI in it a, a reason why there were so many people in the group? I don't know. Maybe Have we all maybe AIed out a little bit?

Absolutely. You know, and I'm not downplaying AI or bad mouthing ai, but it's refreshing to have a discussion that's not necessarily leading with ai. Um, you mentioned Postgres, right?

Yeah. Now Postgres is sort of one of the best kept secrets of, it's not a secret, but you know, it doesn't get the de the, the the and what it deserves, the acclaim it deserves. It really has become an engine, you know, and it started, I'm not, again, I'm not taking sides, but when my, when Oracle bought MySQL mm-hmm.

It set Postgres on fire. Fire. Yep.

Yeah. And, and since then, you know, you have, there are several different Postgres, uh, providers now, inversions and some pure open source, some open core mm-hmm. What have you.

But it really has become sort of the database or record, if you will Correct. For, for a lot of these large, large, you know, hyperscale kind of environments. So, you know, that's something that, um, you know, I think a lot of our audience realizes it, but there it's worth repeating.

It's worth saying out loud. Yeah. What I can attest is the performance issues we saw prior years when we were running MySQL with lot of optimization hacks and everything else since we did the migration platform just scaled up and has been performing awesomely and we were apprehensive, like is it just one of those things that we have to do it with no gains, but we are seeing real benefits of this migration on top of it.

Those of you who are running Artifactory do consider using direct downloads. It just is cherry on the, uh, cake and will give you more performance. Really?

Yeah. Direct downloads from artifact. Yeah.

Just my two sense and progress, um, means it has become, um, such a good database and persistent choice that in the recent years, at least in the last five years, anything that we have that my team has built in-house has to Postgres as the persistent solution. So the developers who are on the ground writing a lot of code day in and day out, it's their choice to do relational databases and persistence. So plus one to what you said post has become that thing where, where probably Oracle of the last decade or, uh, so like, but but it has become that like Database.

No, it is, it is. So I've been in this world a long time. My SQL was the standard Yeah, yeah.

In SQL database because, you know, you weren't locked into a vendor and it was, you had the community developing it in essence. And, and maybe it was, it was probably still a, a small handful of people who were contributing code, but you had the community driving the, the, the, the development visions and, you know, feature set requests and, and quite frankly, you know, Martin Mickens and the people who were running it back then. Yeah.

And it was, it was a great open source success start. Yep. Yeah.

So for any project beyond the con core contributors, it's the ecosystem around it. Yes. Like the different use cases.

Oh, I, I, hey, just a conversation just between us, like, hey, I'm trying to build this app and this scale. Go use Postgres. Here's what you should do.

Yep. So these things which we just take for granted, basically make the, somebody wrote a blog post or we came in and spoke about its success in just specific to artifact. These things do add up and Absolutely.

To, or as technology more successful Now, you know, you look at Swamp Up and I've been to a lot of Swamp Up in many ways. This is the company and the conference that Artifactory built. Yes.

Mm-hmm. Right. That was the Acorn.

Uh, there's Oak Tree grew from, but there's more to Jfr here. There's more to swamp up than just Artifactory. We saw today, you know, this week we, we saw, uh, uh, uh, JFR fly the, the agent AI repository.

We saw the AI catalog. We saw a lot. Yeah.

How does that fit in with your mission at Adobe and your, I maintaining your, uh, IDPs and, and your developers? Are they using these new tools yet, or you think they'll want to use these new tools? Uh, go ahead.

Going off to you. So, Um, we just started, uh, we were locked in Origins because of the database migrations. We just started opening up.

We are current and now we are in a position to start exploring and getting benefits from the machine learning repos coming in, or the catalog. And certainly the agent take workflows. We, we are going to go in and see where all working with Jfr, uh, our partners, see how we can tap into these new offerings that have come in.

And let's not forget about the SBO m uh, AppSec, uh, offering that has come in too. So we, we are going to go and explore these. No, we are very interested.

And so I think, um, there are two sites to it. One is as a platform group, whatever we are offering our customers, we want to make those capabilities more urgent tech. So we, between Vishal and myself and our teams, we need to build more agents.

And the capabilities that kind of were, uh, published in this year's swamp up some of that we are very interested into, um, looking at and see how that can help. Uh, the second side of it is, as our product teams are taking that agent journey and bringing in agent features into the products product like Photoshop, illustrator and several others, they need a platform where they can run agents. So we are solving that agent platform problem as well as we are building agents for ourselves, which would let us kind of expose our capabilities back to, so on both sides, the, the, the, um, the announcement that came from jfr, the excites, the Excite Us, be it AI catalog or any other feature, I feel that can really help us.

Um, so we will go back and start looking at them and see how we can factor them into our, um, our use cases. I love it. You know, um, people don't realize about Adobe.

You know, Adobe is, look, it's kind of a Blue Blood, a royalty name in, in the software world, we all came up using Adobe products, whether it's for video or graphics or PDF or what have you. But I don't know how many people out here really realize Adobe is a company that's very transparent about how they build, how they secure what they do internally, like their their own best customer in some ways, right? We we're gonna show you what we think are best practices and you should feel free then to use them, right?

We, we've already paid the idiot tax in some respect, learning these things, and we're trying to save you from paying that tax going forward. I don't think they get enough credit. I mean, you make great Photoshop, you make great Acrobat, but really being a good, a good community member Yeah.

Right. In the software industry is, is commendable. And, and I, you know, congratulations to you not just that you two, you but the whole organization.

Yeah, yeah. For the, for the way you do it. It's, it's really, I wish more companies we're like that.

Absolutely. No, I think, uh, this has been a practice in the company since long time before we began our stint here. And we are just continuing that, right?

So our leaders tell us, like, whatever we learn, it's our responsibility to go and pass on to the community. And we are doing it. Vishal and his team did extremely great job of this really difficult migration.

I want to thank him and his team and all of the people who really did the work on the ground and his team is coming in and sharing all of that with the community. And it, it, there are a lot of interest. I'm sure other teams are going to do the same.

And, uh, this is going to happen A lot. You've well received. Yeah.

Where's, where's, where's the next conference you're presenting at? I haven't decided yet, but we'll identify something soon. Yeah.

Our group is, um, going to CubeCon. We have some presenting. We will Be, we'll be a CubeCon going live.

com site. Yeah. To find out more what you're doing with platform and how you're building it out.

Yeah. Because this is, you know, DevOps, cloud native platform engineering, these are all just different pieces of today's software factory, including ai. You know, AI plays on all of them.

But yeah, It, it, it just basically, uh, compressing, uh, I was talking to one of the, uh, conference attendees. What AI or generated AI in particular has done, it has flattened the learning curve, like getting a particular sector or a piece of technology or coding for that matter where, which is what Shibu and I are closest to the getting a prompt in and start learning about, or basically porting a technology which was written on one technology to another technology. It's not completely solved, but it certainly is easier.

You can understand legacy codes better. All of these have intangible, uh, benefits and that's why we'll see more software coming faster at us, which basically puts stress on all the underlying plumbing or the CICD supply chain. That's where the, the platforms had to come up to basically, um, deal with the new throughput of, um, innovation coming in and making sure that each of the right customers.

Absolutely. Gentlemen, I want to thank you. Thank you Micha.

Pleasure. Shibu. Thank you.

We're gonna wrap up our, our swamp up coverage. I hope you've enjoyed it. We will be back.

Well, I guess we'll be back tomorrow with more texture on gang, just not out here in beautiful Napa. But until then, this is Alan Shimel. On behalf of Jay Frog and all of our guests, thank you for watching and staying with us and, uh, we hope it was valuable to you.

We'll talk soon. Bye-bye.