In The Arena by TechArena - The Cutting Edge of Data Center Infrastructure Innovation with OCP’s Steve Helvie

Episode Date: April 24, 2024

TechArena host Allyson Klein chats with Open Compute Project VP of Emerging Markets, Steve Helvie, about the proceedings in Lisbon this week and how OCP is helping to shape the cutting edge of infrastructure innovation.

Welcome to the Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome to the Tech Arena. My name is Alison Klein, and we are recording this week from OCP Lisbon. It's a fantastic conference, and I'm so delighted to have OCP's own Steve Kelvey, VP of Emerging Markets. Welcome back to the program, Steve. Thanks, Alison. Appreciate it. So we last talked in Prague last year, but why don't we just start with an introduction of your role at OCP and how it's shaped in 2024? Sure. Part of my focus within the foundation is emerging markets. So that's not necessarily geographic, but it's more technology.

So as specifications come into foundation, they need a place to go to market. So I help those members that are making those contributions in the supply chain around that go to market through various programs that we have around certification programs, data center operators that may be able to exhibit those type of technologies, as well as full rack systems integrators that then take those components, move them into a full rack solution and out to market. This is such an interesting part of OCP for me because it's not just setting the specifications on form factors or some of the things that are associated with standards groups, but really working as a community to take technology and take it from

ideation into broad market adoption. And there's so much that we're covering at OCP and Lisbon this week. I just was in the keynotes. It's amazing how many new technologies are coming to bear. I wanted to talk to you a little bit about infrastructure build-out in the data center. You've been such a critical force in ensuring a choice of our offerings in the marketplace. Can you talk a little bit more on that value map that you just charted, how you're looking at ensuring that there's a choice of offerings in the industry and what do you see that works as we introduce new technologies into the stream? There's really two aspects in my mind.

One is on the contribution side. So one of the big differences we've seen is when OCP first started, you would have a monolithic type of spec where you would have one supplier writing a specification and then taking that specification to market. What we have seen over the last year to two years is a proliferation or expansion in the number of companies working together on one specification. DCMHS, or the Data Center Modular Hardware System, is a great example of that, where we have multiple companies, almost eight to ten companies,

working together on one specification. What that does is it really enables the supply chain toward the end of that specification. So when the specification is done and approved through the technical committee, then it makes it very easy for these suppliers to say, yes, I comply with that specification, and then I can move that particular product out to market. We see the exact same thing happening in the cooling side, where we have multiple manufacturers adhering to, say, a connector, and that really does accelerate the supply chain. So I would say that that's one big area is that we have simplified the contribution process

and the specification process on the one side. And on the second side is identifying those suppliers early and helping them understand the channels to take that to market. Because ultimately, that's what we want. We don't want a library of specifications that sit on a hard drive somewhere. We want products to market. Now, you came up with OCP Ready. Tell us about that and what is the operator response? There's two versions of OCP Ready. The first was launched in 2018, 2019, right prior to COVID. And that was really designed around helping data center co-location operators understand

what does it take to deploy an open compute rack, because it is slightly different in the way that it may come fully loaded, it may be a little bit taller, the ramp heights, et cetera, that needed to move a rack in. So the entire guidelines were shaped around the rack. Fast forward to this coming year, we have now with to a second version of those OCP-ready guidelines. And they're co-authored by Google, who's co-leading the project. And this is to help co-location operators and data center operators understand what's needed for hyperscale deployment. So it's well beyond the rack. Now, what are the hyperscalers needing in these facilities

to deploy OpenTQ hardware? And this could be everything from the parking lot and how big the parking lot needs to be, these trucks around, to the data hall build-out. We've even expanded it to facilities that may not quite be ready yet or being constructed yet. So we have core and shell opportunities all the way through data halls to build up.

So now we're taking, because the demand is such that people need to know when capacity is getting ready to come online well before, say, even the data hall is built out, we've expanded those guidelines to catch folks earlier on the cycle so that when they're ready to build the design, that it is OCP ready from the beginning. We're in the early stages yet. The majority of this work will take place in Q2 and Q3, but the guidelines will be covered again here at this Lisbon event. That's fantastic. Now, I'm going to switch topics a little bit.

Security is always a top operator of concerns, and rightfully so. OCPs have been investing in this space. Tell me about OCP Safe and why the organization decided to focus on firmware. Well, firmware is pervasive in every device that's in the data center. So that's ubiquitous to the data center and it's a natural fit and a natural extension for what we do within the community. There are pervasive issues now in firmware audits that take place. So Google and Microsoft came together and started to lead this effort where what's happening in the industry today, Alice, is you have the same type of audits being performed on devices at the request of different customers.

So a device vendor may have to do the same type of audit for multiple customers over and over again. So there's some redundancies in the market that increases expense and time to market. So what we hope to do with the security review framework, SAFE, is to reduce that redundancy and then also help provide an up-leveled conversation when it comes to security. Because while the hyperscalers may have the resources to go do these audits or have them conducted by a third-party audit firm, there are many in the industry, large enterprises or folks that may not have the resources to go do a security audit. So we have the redundancy in audit, expenditures, and cycles from the device vendors and the end customers. And then you also have the downstream impact of people being able to leverage the work that's being done in the community by providing these short form audit reports. So on the device vendor side, you have a device vendor would work with an approved OCP security review provider or an SRP. And these

third party audit firms work directly with the device manufacturer to produce a short-form audit report. These short-form audit reports will be posted on OCP around the device for anyone to download. And it does give everyone that sense of comfort that OCP is starting to address. We're not a security audit firm. We don't claim to be. That's all being done by the approved security review auditors. But what we do hope is that with more firmware being through the OCP safe program, it starts to up-level the conversation. We increase our posture around security overall.

And OCP provides the perfect governance framework for allowing people to work in a secure environment like this, rather than having two or three main companies have their own preferred SRPs and do their own audits, et cetera. So again, a good example of community coming together, identifying a common problem, and hoping driving some inefficiencies out of the market. I love that. What do you think is next in this space as you look at how you can address the full challenge for operators?

Being safe itself? Yeah, exactly. Yeah, it's right now the very first product that's been through the OCP SAFE program has been hard drives. And so the majority of the interest now has been hard drives. However, at the OCP Regional Summit this week, we'll have AMD on stage talking about how they're running some of their GPU offerings through. So I'm hoping that that's the beginning of a lot more of those type of devices coming through SAVE. And of course, hard drives are a very natural fit, but I anticipate OCPC hopefully expanding

across multiple, but AMD being one of the first, again, on the GPU side. That's really exciting. And Tech Arena audience, watch this space, because I have a feeling that we're going to be seeing more about OCP Safe in the months ahead. And it's relevant for the European market, for sure. Oh, yeah, for sure. Now, we're going to the edge, and specifically the edge of the RAM with Evenstar. You know, it's interesting.

When OCP first came out, I never thought we would talk about Open RAN. I mean, that's such a different topic than, you know, Cloud Data Center. But it's so cool to see OCP venture into this space. Tell me about Evenstar and why that's such an important initiative in terms of enabling the edge. It's a good example of emerging markets for OCP. I'm learning about Open RAN 2 a lot over the last 18 months. It's important because it's a natural fit for what we do within OCP

is building multi-vendor supply chains around a piece of hardware. As Meta contributed their specification, their design files, OCP was a great fit for taking that specification and start to build a community around it. We see, as most people have indicated, the radio unit itself is one of the more expensive proprietary areas of the entire network. So with our ability, hopefully, to disaggregate that

and drive down the cost model through open source and through open collaboration, we hope to provide a radio unit that is applicable for a multi-use case scenario, so a family of SKUs, not just one radio unit. So the meta specification is for one type of radio unit. We're trying to decide now, at the beginning of this, including this week, is to continue to get feedback about what's good about this specification,

what can we use out of it, and then how do we take that and then build a family of SKUs around it. It could be private 5G, it could be a type of radio unit that you would use in an emerging market. So it is really the beginning of a foundation of a community. So you're at the beginning stages of this open radio unit. And there are some people that are very bullish and think that this is going to go well. And there are some people that are pessimistic saying that, look, there's

got to be a lot of interoperability. And then the more disaggregating you get, you drive up the cost, not down the cost. So we'll see. But we are pretty optimistic that we can find a really good use case and family of SKUs for an open radio unit. And this particular area is ripe for an open source or an open box. Now, this is such an interesting point because I've been following VRAN and Open RAN for years. One of the questions that I have, and maybe you've come to some determinations, is why is this such a hard nut to crack in terms of cloud native technology and open technology? Do you have any perspectives on that? Only from the perspectives

I get from listening to others. So there's a couple of things that have come to mind. One is it's very incumbent driven. So the incumbents have a large advantage and a lot of value add in their proprietary solutions. Those proprietary solutions come at a high cost. The other is making it, that last point that I raised around the interoperability or when you start to break something apart, there's a chance that getting it small enough and interoperability enough or interoperable enough that you end up driving up costs and not down costs. So those are the two areas that I've seen. And consolidating demand. So let's say that you have a great radio unit. You still need to have a manufacturer out there that has the demand ready to warrant building up a production line.

So what we've run into in the past is there's just not a consolidated enough demand for one of the large manufacturers to say, well, I'm going to move off of this. I'm going to start doing something on an open computer. So those three areas, incumbent, can we get the cost down, and then can we consolidate demand? I think number two and three, when you get the cost down and the interoperability and the testing piece of it, the third will follow pretty easily. That second one to me is the bigger issue. Now, we've covered a tremendous amount of ground between OCP Ready, OCP Safe, Evenstar, tons of topics in terms of areas where OCP is helping shape the future of technology. I'm sure we've got folks listening online that either want to learn more about what

OCP is up to or hopefully get involved in the work groups and the movement forward with these technologies. Where would you send them for more information and to engage your team? I would send them to three places. First, don't struggle finding all the information that's on our site. Feel free to email me directly at steve at opencompute.org. So I'm happy to help navigate those resources. Second is our website. There's a lot of information on there, a lot of mailing lists, a lot of ways that you can start to get engaged in the specific projects. And then third, our social media channels. So our YouTube channel, our LinkedIn channel are great to see the post and our community at work. All of the sessions this week will be recorded and posted online. So you can go onto our website and go to

past events under events section, and then jump to those sessions. And then from there, you'll start to see where you may or may not be able to fit in. But we encourage everyone to at least take a look. Don't jump in right away. Understand the community. Figure out where it makes sense for you to add value to your business through an open source community. Steve, it's always a pleasure. Great seeing you again. Great seeing you too.

Thank you for taking time out of your busy week in Lisbon. Absolutely. Thank you very much. Thanks for joining The Tech Arena. Subscribe and engage at our website, thetecharena.net. All content is copyright by The Tech Arena.