Equinix’s Infrastructure Solutions with Steve Martinelli and Jeffrey Sica at KubeCon Paris 2024

This is Textron tv Everybody. Welcome back to Coup Con here in Paris, 2024. Day three more fantastic interviews.

You know, we've told you we've got some amazing guests, and we do, and it just gets better and better all the time. We're gonna talk about, uh, projects, open source projects, and what some of the needs projects have, and how can companies collaborate, contributors, collaborate, uh, provide resources and things that, uh, projects need. I'm joined by a couple great folks, Jeff and Steve.

Jeff, you introduce yourself. Sure. Uh, my name is Jeffrey Zika.

I am the head of projects at the CNCF. Great. Steve Martinelli.

Hey, I'm Steve. I'm a director of developer advocacy over at Equinix. Uh, and as part of that mission, I also, uh, help out with our open source project program.

Very nice. And you both have been kind of involved in open source for a big, better part of your careers, right? Uh, yeah.

Just about You wanna go first? Yeah. Um, I've been working with open source projects for at least 20 years, and I was a Kubernetes contributor even way back in 2014.

So, Early days. Early days. So I, I have, I, maintainers and contributors are my people.

Like I hang out with them a lot more than, you know, other portions within the foundation. So like, I feel like I have a pulse on what project needs and what maintainers need. And I mean, I'm passionate about the projects.

And you're a staff member with CNCF. So what does that, what does that entail besides everything, whatever needs to be Done Besides, besides what fire to put out. Yeah.

Um, I know a lot of the maintainers in the projects and I know a lot of the foundation staff, and it really is just trying to interface the two and make sure that everyone is coming out and getting what they need and having positive outcomes. And your name's Jeff, but your handle is Gt. So gp.

All right. A little over 20 years ago, a friend of mine misspelt my name on a OL instant messenger and it stuck. Be careful.

Those mistakes become lifelong, uh, decisions. Yep. Sometimes.

Steve, how about your story? Uh, you know, it was 2012 and I, BM was doing its usual reorg and I landed in this open source team. I joked that I was traded for future consideration.

Ah, uh, 'cause it's not like my team got something in return. It was just like, I just got a kind, A banished open source or something. Yeah.

Banished to open source land. And this is 2012, right before I was going on vacation too, of course. Uh, but honestly, it ended up being, uh, such a great time in my career.

It's my favorite five years. I was working on open on OpenStack, just doing upstream development every day for five years. It was a wonderful time in my career.

I look back on it very fondly. Uh, you know, I ended up being a maintainer on probably a half dozen projects and a project lead for one of the, for the identity and authorization, uh, service. Um, just a great time.

Got to travel the world, got my first, uh, gotta go to my first conferences, speak at them, uh, you know, great times. And ever since then I've been smitten with open source. It's always been a part of me, so I'm really to shape that career tra trajectory.

Right. Yeah, it really did. It was great for that.

Excellent. Well, you know, we always think about contributors of open source projects, doing some of the things you guys have done. Yep.

Writing code, debugging, you know, all the, all the activities, but you gotta have an environment to run those things on. Obviously that's a, a big plus of Equinix being such a, a strong sponsor Yeah. Of projects.

What are the, is it just you, here's some hardware, go have fun? Or what, what do you have to do to support projects so they can be productive? Yeah, so it's kind of a big question there.

So you can imagine with 250 data centers all over the world, we have a lot of hardware. Um, and you know, as part of our Equinix metal offering, which is our dedicated cloud, where you kind of have the best of both worlds. You have the, um, the performance of bare metal.

So you can get right down to the silicon as much as you want, but you're also getting the best practices of cloud. So you can provision those servers, uh, you know, in about two minutes you select through our console, go select the operating system there, select the server configuration, the metro, and all of a sudden two minutes you've got your bare metal, you BUN two server. Um, so, and I forget where I was going with this one here.

No, no, we're, so what are the kind of the things that projects need, right? There's compute And storage, so there you go. So with that, we give them the compute and it's really just comes down to the, what the projects need.

So the projects will often need, um, you know, they wanna do their development, so they want to end up doing, they want to end up, uh, incorporating their environmental servers in their CI ICD pipelines, right? Yep. So we'll see them use, uh, Equinix metal for, again, CI pipelines, doing builds.

Uh, um, very common pattern actually is them doing performance testing and actually like integrating performance testing into their CI builds. Um, I mean, over 180 projects, we have a bunch of networking projects, CNI and service mesh projects. And they need to be, again, as you said, as close to silicon as possible, because they're not just running in, you know, cloud providers.

They're also running in data centers, and they need to make sure that the performance is solid, it will scale out. So Equinix gives them that, and it works out really well. It sounds like that, uh, fabric of 250 worldwide data centers has gotta be a big plus too.

Like, ID need to do distributed apps, not just in a data center, but across geolocations, maybe different kinds of hardware, whatever it might be. Oh, yeah. Uh, actually one of my favorite stories, uh, out of our open source partner program is probably Alpine Linux, uh, which is the base image for a bunch of containers out there.

Uh, and they're a big, uh, consumer of Equinix metal, and we're happy to support them. We give them access to both X 86 and ARM servers, and they're able to not only, uh, build the different images and their daily images and all that fun stuff, but they also use us to distribute, uh, those images also. So it's great.

Talk about the importance of arm, because that comes up more and more. I mean, I've talked with so many people about, not just at the Edge, but in the data center. Yeah.

So it's funny, you know, I saw every time I read about arm, it's always in terms of performance and efficiency. Mm. And to me, it just, I, you know, it just seems like even on our MacBook, right, with the M ones, uh, they are just, they just seem to be a step ahead.

Um, I can't say much more than that. I mean, have you seen a lot of demand for it? Uh, significant demand.

A lot of our projects actually try and cross compile, so they're hitting a MD 64 a Arch and arm, uh, and then, you know, other, you know, less common platforms. But I mean, at this point, A MD and arm, well, A MD, like 64. Yeah.

Um, between that platform and then arm, they're pretty much treated like first class citizens, uh, with the projects. So, popular. Popular.

It's interesting. I was talking popular, I was talking with somebody from Arm about, you know, just as an aside, how much do you think Apple moving to the M1 a, the ARM architecture, even though it's not Arm's product, um, has really opened the White Oak. 'cause now we need all that stuff to run locally.

A lot of times didn't hurt, right? Yeah, It definitely did not, not hurt. All of a sudden all those developer tools had, uh, arms 64 builds very quickly.

Plenty of how that Happened. Yeah. Equinix actually had a big part in helping out with the Works on Arms Project and actually helped a bunch.

This is before my time though. Uh, shout out to Ed Eddy for doing a lot of the legwork here. Yeah.

Who still works with GFI today? Um, we worked a lot on the Works on Arms project, um, and just getting a lot of different, uh, developer tools and developer libraries up to, you know, being, uh, armed 64 compliant. Let me, I, I just want to like, not only plus one and say thank you, ed, but like, he was actually going to every individual project asking them, how can I help?

How can we enable, like, provide arm enablement? Do you need resources? Do you need engineers?

It was, it was a herculean effort done by a single man. Yeah. Shout out to Ed.

Shout out Ed, that's the star here on this show. Thanks, ed. How about, so why is bare metal, why is being able to develop tests, performance test on bare metal, so important as supposed tossing it into some whatever virtualized environment, cloud service, and kind of taking it from there?

I think it's a lot of it comes down to performance. A lot of it comes down to needing control. So, you know, when you're doing driver specific or hardware specific things, you want access to that hardware.

You don't want to virtualized, uh, you know, kind of a virtualized version of it. You want to actually test on those things. So, you know, for operating systems, yeah.

They kind of want to access, they need that control. So they need access the hardware. Um, what are you seeing With projects?

It is, it is, unless you've run into this issue, you might not think about it. But with cloud providers, you're still fundamentally usually running on a virtual machine or in some hypervisor, some not directly silicon, you know, environment and projects run into issues where they might break the hypervisor or they might run into some compatibility issue with a certain version of what the, you know, hypervisor is. And there's no real other way to debug that other than, you know, either try different cloud, oh, they might be, you know, set up the same way or go to bare metal.

And then you're essentially eliminating all the, you know, middlemen and then going, okay, it's either my code or, you know, it's the hypervisor. And then they, they at least know they need to debugged somewhere else. It's eliminating yet another variable.

Yeah. And you're doing these things. How about the need to deploy on bare metal?

Is that increasingly a need projects have, because you hear about Kubernetes on bare and metal and things like that? Yeah, for me it comes down to, uh, use case, I guess. Uh, we're seeing a lot of use cases for bare metal.

Um, you know, geez, when it comes to CICD, when it comes to, uh, backup and disaster recovery when it comes to streaming, again, really takes advantage of the latency and performance there. Mm-Hmm. Um, online gaming, that sort of stuff.

We see a lot of use cases around there. You know, you're not gonna use bare metal to deploy whole world. You're not gonna use it for your sample app.

But in cases where you want like, um, data sovereignty, uh, you know, or performance and latency, I think bare metal was a great choice for That. Makes a lot of sense. So off script.

Sorry. How, how much like AI deployments? I'm sorry, I to bringing No, I was going AI next.

Oh really? Let's talk about where AI fits into this Field. Like, I'm just, I'm just curious like how much in the last year you've seen, like, in terms of uptick.

So, you know, what was it, uh, what, what, what did the Nvidia CEO just say like yesterday or two days ago at data centers or AI factories or something like that? I thought that was really cool. I'm like, yeah, data centers are getting a second wind here.

New Tagline. We just got that up on our website five minutes after He did that. I like, what a great soundbite there.

Um, and so yeah, it, you know, Equinix just announced this partnership with nvidia. Uh, you know, I'm still maybe one step or two steps removed from all that stuff. Uh, but yeah, it's like a ton of demand.

It, it is. I mean, that seems to be a shortage of getting video cards, shortage of getting chips, shortage of, okay, now how do I get that in, in the cloud and service providers? And it's, it's funny because there's such limited supply right now, um, and it kind of makes sense if you're, you know, renting a GPU from a cloud, you're probably probably playing, probably paying an astronomical price for that per hour.

Uh, if you actually bought that card and put it in your data center, you're probably saving a bunch of money in the long term. Um, so, you know, it, it, it's a big supply and demand problem right now because people can charge basically whatever they want. So it's, it's an interesting problem to be in.

And I think data centers are well positioned, uh, to take advantage of this. It seems like it's, you know, AI isn't a new field, but the budding need for so many GPUs and specialized chips, that's hard for someone who's building a data center to know what, you know, you build to the capacity you're gonna need in five years, not what you need today. So how do you know what you're gonna do and in your environment you can build to that evolving need in the market, you know, staying whatever steps ahead of to be able to do that.

So when the demand's there, you can satisfy it. Yeah, exactly. Um, yeah, shout out to our folks working in our data centers who are actually like the technicians and the operators.

They're, they're handling that and the folks actually planning all the server deployments there. So they do a phenomenal job. Yeah, they do.

Thank God they're there. We wouldn't be here without 'em. Right.

A lot of things would not, a lot of things would be breaking right now without them. Yeah. Don't anybody, but they're doing God's work here, you know, this is important stuff they're doing.

Right. Good things. What's, um, what's kind of on the horizons for projects and what people are working on that you're, uh, kinda grabbing your attention?

Yeah. No, no, no, no, no, no. So it's, it's kind of interesting, you know, I I of course throughout the AI thing, but we also are seeing, you know, AI related projects starting to be submitted to the CNCF mm-Hmm.

We have Kate's GPT, which, you know, lets you do LLM analysis of a cluster and kind of aids in debugging. It's meant to be as an SRE with superpowers, I think is the tagline. Ooh, I like that.

There's another tagline. Oh, yeah. SR um, but also, you know, you look at what the CNCF provides applications, in my mind, AI and LLMs are kind of the application layer.

So what does cloud native have that we could enable LLMs to perform better? So let's talk about telemetry. Uh, there was a project that was submitted, didn't get in, but they were asked to re reapply in six months called open LL imagery.

It's a mouthful. Yeah. But the, yeah, EIO.

Okay. All right. But the Whole purpose of it was getting out telemetry data from LLMs, you know, not just like GPU information, but you know, tokenization, All of that.

Absolutely. That sounds like a phenomenal, We're gonna see, in my opinion, we're going to see a lot more of those kind of projects coming in. Now what about the projects that are already here?

Believe it or not, and this is what I was trying to allude to. We already have projects asking for GPUs, and I'm like, what? Why No.

Should be asking why. Well First, what then, okay, what is your use case? Uh, still waiting to hear back.

Yeah. I'm not sure if they want to spin up an LLM and try and integrate it with, you know, their project, but It may be experimentation and like, it could be a specific use case too. I am, I am here for this.

I want to enable this, but I also want to make sure that it's kind of used responsibly and they're not just wasting like an a 100 for a month People wasting resources on cloud. You don't say, I mean, Life are users at anywhere in the stack, you know, they're gonna ask for what you never expect. Right.

Yeah. Well, and also, you know, it's very easy to just be like, I want this now. Click you done.

Well, as soon as you say, I don't need that, you'll need it tomorrow. Right. So tell me, you know, I don't wanna make it, you know, a big, big huggy thing for, for Equinix, but what's it like working at a company that really is, you know, you're supporting open source projects in a big way and you know, you've been a contributor, so you know what it's like to be on the other end of that.

Oh yeah. Resource starved projects at times. Uh, you know, it's funny because, uh, you know, the really, really popular projects get a lot of the contributors, get a lot of the hardware resources, get a lot of the, you know, the fly by contributors get a lot of maintainer support.

It's all the other projects, which probably outnumber them, tend to one that are like, Hey, can I have some help too? Yeah. What about me too?

Yeah, I could use a little bit of help over here. And, uh, you know, it's, it's just as being on the other side there. And now I'm able to actually help out also in terms of providing, uh, credits and infrastructure for these resources too.

But, you know, that's just our way of, uh, contributing and helping out projects you can contribute with whatever ability you have, whether that's documentation, whether that's hosting a meetup, uh, whether that's, you know, going to a KC a more local KCD if you can't come here, uh, there's ways that you can contribute open source, uh, more than just lines of code. Well, and I gotta believe it's, it's valuable having a role like yours that can maybe do the interpretation, if that's a good word. Uh, what are projects asking for?

Why do they need it? It's just not giving them free resources because it's nice to have, right? Yeah.

Here's why they need these things to use cases, whatever it is, and then translate that to folks that may not know as much about it. Yeah. Yeah.

It's, uh, yeah, it definitely gets fun being a translator. J Jeff Jeffy, if you had any advice for projects, maybe like folks that are submitting one, what, what are some of the capabilities they, they have access to that may not be, uh, aware of yet? So the short answer is we have a website that lists everything because we have a lot of things that we offer.

The simple answer is we will try and do whatever we can to give projects what they need. Uh, we certainly have a full list of, you know, resource providers that we offer, like Equinix services we offer, like helping them in the project pavilion or events, marketing swag, et cetera. But really that, that's not like a finite list.

We, you know, we'll provide that. And then at the bottom of that website that I just mentioned, we say, and if you need anything else, please reach out to us. We are happy to talk because I can't predict what a project needs in three or six months.

Like if, if we try to prescribe too many things, if we try to put projects in a box, we're gonna stifle their growth. And we just can't do that. Like the whole reason, look at all of this.

The whole reason we're here is because we kind of let projects go and do what they're best at, which is innovate and hack on cool stuff. And then, oh, hey, 12,000 person convention. Hi.

Great. Well, um, thanks to both of you. Thank you.

Contributing when you were developing and contributing with what you're doing now in your roles. And, you know, it, there's no one person could make this all happen. It's all these different roles and, and contributions that everybody makes and the belief and sustainment of that.

So thanks guys. I appreciate it very much. No problem.

No problem. Great talking with you. Steve.

Jeff. Yep. Thanks a village.

Thanks Steve. Absolutely. Thanks both to Equinex and to the CNCF team here, having both, uh, Jeff and Steve here with us.

So, like I said, amazing, amazing people. We're talking you, we really talk to the people who are helping make this happen. And there are so many folks playing vital roles like these two guys are.

So stay tuned. We've got more great interviews coming right back up here from C**n in Paris. So same bat channel, same bat station.

Don't go away.