In The Arena by TechArena - Data Center Innovation with OCP

Episode Date: May 18, 2023

TechArena host Allyson Klein chat’s with OCP’s VP of Market Intelligence and Innovation, Cliff Grossner, about the challenges facing data center innovation and how OCP has transformed to drive deeper partnerships with the industry and broaden its impact on data center 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 today I'm delighted to be joined by Cliff Groszner, Vice President of Market Intelligence and Innovation for the Open Compute Project. Welcome to the program, Cliff. Well, thank you, Alison. It's great to be here, and I look forward to a great discussion. This episode is part of a series of episodes about what's going on at the Open Compute Project, stemming from the regional summit in Prague. And I'm so glad that we've finally gotten a chance to catch up with one another, because

I want to talk to you about what it means to be the vice president of market intelligence and innovation at OCP, and what your purview is for the organization. Well, it's funny you ask that because there was no such role as market intelligence and innovation at OCP until I joined. And part of what we had to do is figure out, gee, how could I be useful to help move the community forward? And so I took on a number of challenges because when I joined the OCP, it was extremely well known in engineering circles,

but not in the rest of the world. So what I told the OCP is, because I have been an industry analyst for more than 10 years, I can now turn around and use the analyst channel to bring knowledge of the OCP back to that channel and out to the world. I explained to them that I thought the OCP could enhance itself by being more present at third-party events and build alliances in a bigger way than it had done in the past. And since I joined OCP, we've definitely moved the needle on the analyst engagements.

We have a much larger presence now at third-party events than we had in the past. And we continue to grow the alliances. You may have heard that we announced a pretty big alliance with Linux Foundation last year. And we've got a couple more coming up this year that I can't just talk about. One of the big ones we announced last year was with Jetpack. So you can see the remit. The other

element around market intelligence and more the innovation side is coming in and helping the Open Compute Project take its future technologies initiative and future technology symposium to the next level. That's a part of the OCP that looks at bringing in communities from academics and industry researchers to complement what you have in our OCP projects, where those are people working on today's product and tomorrow's product. What you're describing is incredible. And it talks about the change in the OCP. You know,

I think I worked on the launch announcement of the Open Compute project back many years ago. And, you know, I think that anybody in the industry that was involved at that time would say great organization for the large cloud service providers to share what they're doing and have reference architectures that could be scaled, but maybe not as relevant to the everyday data center customer and maybe not as broadly proliferated in terms of impact. But what you've described is completely different. So why don't you just take us back into what has changed and how has that changed during your tenure at OCP in terms of that broader engagement with different organizations and different types of customers? Right.

And in fact, I cannot take credit for all the 10 years of evolution of the OCP since I'm just coming up on my second year here. But if you want to laugh, I actually remember your announcement because I remember reading it as I was in the industry at that time and going, wow, how cool and how interesting and how groundbreaking that could be. And you're right. The original incarnation of the OCP was to create open hardware specs built by some of the large data center operators so they could get the equipment they wanted. They could come together in a collaborative way, write specifications, and have vendors that were willing to build to that spec. And today, that remains part of the core function of the OCP. And the original tenets of the OCP were around openness, which is still existing.

But it's evolved considerably to today. Probably if you ask me what's the most important part about OCP, I'd say it's collaboration. The community coming together to collaborate, either to work on a design that they all can share to the betterment of the market, moving all of their companies forward, or some standardization. And the OCP is not a standards body, but we do work on a lot of standardizations that help move the industry forward. Some examples of that might be the alliance we struck with JEDEC late last year, which is something we probably would not have done two years ago, where within the OCP, the community came together

and specified a schema for describing something called a chiplet, which is a small die that would be combined with other dies to create a system in package, a new way of creating a system on chip. And this is potentially going to revolutionize, in my mind, it's a big disruption that we're going to see in terms of how silicon gets built. What's the supply chain for silicon?

Of course, we'll still have the big movers and shakers. We all know their names, but many of the smaller suppliers or smaller companies can now build, design, and sell chiplets to other companies that can combine them into a system on chip. But to do that, you need a standard way of describing a chiplet. What's the size of the part? What's the bump map? What is the thermal properties of the chiplet?

And the schema that I talked about is a standardized way to do that that we've now opened up and through JEDEC, or is going to, or we are going to have that become an international standard. So in the future, all chiplets that get sold in the marketplace potentially will be described using a schema, develop that open compute project as a standardization. Now, just one example of how the organizations evolve from open to include collaboration and from hardware spec to include standardizations. You know, it's so interesting that you bring this up. I've written extensively about chiplet architectures on the tech arena.

It's a passion of mine, being a long-term data center player. You know, the work with JEDEC, the work on the UCIE front is just creating an environment for such incredible disruptive change in terms of data center infrastructure. And I'm so excited to see the industry respond. And especially coming from the Silicon arena, just so excited to see just kind of a renaissance of design opportunity for different smaller players to develop something that may be niche, but may be so important as a smaller element of a larger SOC. It's cool to see that OCP is playing this role. And of course,

OCP can because of the consuming block that it represents in the marketplace. When you think about the future technology roadmap and the efforts that you have there, obviously, chiplets are part of that. But there's a lot of other things that you all are talking about. Can you talk about the broader purview of where you see your focus for tech innovation and what themes folks can expect to see from OCP as we move forward in 23? Right. So let's take 23 and 24 and move our way a little bit. And so in the OCP project space, which would be the next 12 to 18 months,

another very exciting piece of work that's going on is leveraging the protocol CXL, which I'm sure we're all familiar with, and looking at how do we apply that at a system level. And I think the next evolution that we're going to see is memory becomes a first-class citizen in the data center, no longer bound by the skin of a server. And so we'll see compute infrastructure become much more flexible, perhaps dynamically composable, and very much matching the needs of the particular application, dynamically reconfigured.

So after chiplets and composable infrastructure, the next area that's getting a lot of attention is modularity. The fact that we want to think now about hardware infrastructure as individual components, memory blocks, CPU blocks, IO blocks, and have them standardized so that tomorrow when we update, let's say, a server, we update maybe the CPU block, but the rest remains the same. Update the NIC card, the rest remains the same, and they're all standardized. Another, it was moved further out.

We have an initiative that's looking at optical interconnect architectures. We're just getting that off the ground. So today, much of the work that's being done around optics and networking is constrained by the existing infrastructure and the fact that we still have a lot of copper. And people are trying to evolve that, which is great. But we're also saying, let's imagine a world where there's no copper and everything's optical, totally optical, even interconnect on the chip and in the chip. And so what does that world look like? What are the design constraints that get removed and how do we push forward? And we actually have a half a dozen startups that are gathering around to work with us on a

white paper that I'm hoping we're going to produce by summit to talk about what are some of the different options and what does that world look like going forward. So moving to the optical world, moving a little further out, the Open Compute Project is very interested in how it can participate in quantum networking, quantum computing, quantum cryptography. And one of the areas that we see is there's still a lot of, it's the wild, wild west, and there's a lot of room for standardizations. We are familiar with some works where people are trying to deploy the first quantum computing powers, let standardizations needed in the data center so you can connect these quantum computing boxes into the data center infrastructure? And that's not obvious. That's parallel to research that's going on today around liquid cooling, where at the OCP, you're seeing standardizations for the connectors that are used to connect the NIT rack to the building cooling system, the liquid cooling system, and what do those quick connects look like, and how are they

standardized? And so, likewise, we can envision for quantum moving further out that that same concept of standardization can get applied moving outward. If we go even further, an area that we've talked about but haven't done anything with is the idea that eventually we're going to see a new kind of power. It's going to be fusion, right? So you're going to have small fusion reactors that can power a data center. And even though it may sound like that's science fiction and 25 years out, it's not really because we already have projects that within the next 10 years, we'll have facilities up and running and are well-funded in the billions of dollars today.

And I mean, they've got it running in labs, right? So there's a demand. It's in the lab and they're building some real commercial-sized prototypes at this point. So you asked me the question, what does the roadmap look like? This is just some of the things that we think about and work with every day at the OCP. That's really exciting. Obviously, this is all driven to deliver the performance capacity that customers need to run their applications and do so in a sustainable manner. And I've loved to see the sustainability initiatives that have sprung out of OCP.

Again, because of the consumer influence that that represents to an entire sea change of acceleration of industry innovation to address. You talked about the modularity aspects, and we've talked about that with Rebecca weekly on previous episodes of the Tech Arena. Tell me about the industry response to the modularity of design of systems and really focusing on upgrading only the elements that are required

for the application evolution. How has the industry responded to that? And when do you expect to see modular designs like this start hitting the market? Well, they're actually hitting the market now. So it is now. It's something we started a couple of years ago. So it's about the right time we're seeing the fruits of that labor. And the industry is responding very well, partially because the volume demand from the OCP community is pretty large.

So it's large enough that the vendor community will follow. That is, no matter how you like it or slice it, vendors will build what they can sell. Exactly. I mean, another good example of modularity and something that's just hitting the market now is about two years ago, there was an initiative where we recognized or the OCP community recognized that every time a motherboard gets built in a server, it has all these chips that are related to security, making sure the components are secure, the motherboard's secure, boot up is clean. There's no foreign chips on the CPU or on the motherboard or anything.

And every time you design a new generation of server, that had to all be redesigned on the next motherboard. So at the OCP, the community said, no, that doesn't make any sense. We're going to create a new kind of daughter card, which we call our security control module, which is going to, and we're going to define a standard for the connector that would connect that daughter card to a motherboard

and the signal across it. And now every time you upgrade a motherboard, you don't have to update all the security elements. And so we now have several startups that are actually building and have just released their first products. So you could buy security control modules to plug into your servers. And we actually have one or two vendors like, I believe, Ampere is already shipping servers with the connector on the motherboard. That's wonderful. That's really great to see. Another aspect of modularity in terms of sustainability, obviously there's been a

sea change the last year or two with some of the research showing how much of the planet's energy and water and other elements that are being consumed by the data centers. And with the move to AI driving more infrastructures, it's only going to get worse if we don't do something. So I think everyone's realized we're at a tipping point where we have to change the curve. And there are projects at OCP looking at how do we measure carbon emissions and how do we go to new materials

so that we don't produce the same amount of carbon for doing a data center. I don't know if you saw it at the regional summit, but we actually had a wooden rack on display. And it's laminated wood. And so definitely using new technology, but in old material in a new way. And that's potentially going to get rolled out in volumes going forward. Those are the first prototypes that have just come on the market.

That's very cool. That's another example of how the demand there from the community is there to rethink and redo things in a new way, where we can shape that industry going forward. Cliff, it's been great talking to you. One final question for you before we go. I'm sure that there are folks who are listening online that are members of OCP but may not be involved in all the discussions that you've talked about or people who are looking to

join. How do you recommend companies who are part of OCP maximizing the value of their membership and engaging within the community? And then how would you like to engage with folks if they want to talk to you and learn more about participating in OCP? Well, that's a good question. First of all, I did want to point out that you do not have to be an OCP member to join any of our project calls. The project calls are, times are listed at our website and you can just click on the link at the right time and jump on and listen and talk and participate.

And so that's the first way to get engaged is to learn what's going on and just jump on a call. If you need some guidance in terms of, well, what call should I jump on? What's going on on this project or that project? Michael Schill, our community director, is always available. I'm always available. We're happy to have people reach out to us and, you know, provide a, get on a call, give 30 minutes very freely. We'd like to get a briefing from a company on what they're trying to achieve. And then we steer them. We try to help them engage, create introductions to our

project leads with the right, on the relevant projects to their work so that they can engage fruitfully. And there's always just doing what I think a lot of people did this year at our regional summit, purchase a ticket and come on in and see what's going on and meet the team there, which is fine. And the next chance to do that is going to be at the Global Summit October 17th to 19th in San Jose.

And hopefully some of the people listening to this will come and join us. Awesome. Well, thank you so much for being on the program today, Cliff. I was looking forward to it and it lived up to its billing. I really enjoyed the talk. I'd love to have you back on sometime soon. Yes, glad I could help. Thanks for joining the Tech Arena. Subscribe and engage

at our website, thetecharena.net. All content is copyright by the Tech Arena.