 Director of Engineering for a Hyperscale, right? Okay, we're here inside the cube here at HP Labs headquarters, was formerly building one. The first original HP building behind us is the office of Bill Hewitt and Dave Packard, the founders of HP. We're in the sacred spot, Dave, pictures up on the wall here behind us. Getting in the trenches, deep dive, post-event coverage, analysis and commentary. Gerald, welcome to the cube. First thing, Discovery Center. You guys are building all these elements around this big announcement. Dwight was illuminating on some of the generational shifts around possible growth. Great starting point. Use cases that are out there with big data and cloud. Now you've got to go build it. One of the things, component that you're managing, called the Discovery Lab. Okay, Discovery Lab. We saw, we've got a little tour today. Paul gets a little video tour. Yeah, we've got a video out there, but it's based in Houston. Basically what we've been doing with a lot of our Hyperscale customers over the past several years now that we've had that business unit, is bringing in some of the top customers and having them run workloads against our servers. We tune the servers, as Dwight mentioned, with respect to clock rates, the type of components we install. Really purpose-built systems directly for these large, web-scale data centers. So with the Discovery Lab, particularly with some of the announcements we had this morning around the Redstone platform, where the ecosystem is emerging, we really want a place where we can open that up to customers, have them come in and test their workloads on these very different kinds of processors, to go and look at emerging workloads out there as these new extreme-low-energy processors start to emerge in the marketplace and in our moonshot architecture. So what does it mean for the customers? I mean, there's not a lot of customers, are there? Give us the numbers. What's the stats? Can you run us through the numbers? I mean, we're hearing, we're learning, it's a starting point. How far along are you guys with the lab, with the processes? Can you throw some numbers around? Well, so for example, with some of our biggest customers today, where Paul mentioned this morning that a 20,000-unit order is something that they signed a check for now without blinking, we've been running workloads for those kinds of customers on some of our traditional scale-out products already and tuning those products. With Redstone and some of the newer products, what we have right now in the lab are several different SOC partners and suppliers that we've already been running internal benchmarks on. What's the SOC? A system on a chip. So that's from various different companies, including X86 and ARM-based systems. So multi-vendor lab. Exactly. And then we couple the partners in with it because the partner program, the pathfinder program that Mike talked to you about earlier, it's really important to couple the partners in along with the customers and along with HP to really look at the entire solution stack, right? It's not just throwing a piece of code over the wall, hoping that it gets debugged by the community, hoping that it works well for a customer workload. We're really trying to bring that all together in a holistic motion to really optimize around efficiency. So people understand it's a solutions lab. Exactly. It's like an R&D lab as we're here today with HP lab. Well, but it really does start in an R&D lab and my team does R&D. So we bring in the hardware and make sure that it performs well with margin so that we're not exposing our customers to stuff that doesn't work. So we do a core foundation to make sure it works. Then we run benchmark workloads on it and we share those benchmarks with our customers to get an understanding of what's out there. And then we invite them in to run their workloads. How does someone get involved? So take me through the onboarding of a customer. I'm a customer, I'm a big customer. Is there a certain size that can get in? Do they apply? Do they call their sales rep? I mean, they call you. I mean, just take us through just the process. Yeah, so we can have our PR folks send you the link, but there's a link that we've got out there in the Moonshot announcement and then it's by invitation after you've applied to that link. And then what they just get in? Oh, then it comes through our marketing department, then we go size up how many people are available to go and run. They prioritize. They kind of put them in a formal process. Okay, got it. Okay, cool. And so we talked about this earlier, but so the first Discovery Lab is going to be open in January? That's correct. In Houston. And then later in the year, we're going to expand that out worldwide. We'll have remote access to customers where they can log on to the systems remotely. We're doing a lot of that today with our colleagues in HP Labs. That's how we got a lot of the benchmarking testing done already. And so we have a very open environment where they can come lease some time, not lease, but register some time on the systems, come in, run their workloads, look at the power measurements. We've got power meters there all in line on the network. What's the number one thing they work on in the lab? What's the top three concerns that they kick the tires on and play around and really work hard? So for us, it's mostly about stability at a solution level, and then a lot of it's about power, right? When you're scaling these systems out and deploying thousands upon thousands of servers at a time, any variation in power with your workload makes a big difference on the quantity of servers you can go install, the type of UPS and other infrastructure you need to install in the data center. So they really want to size and scope that out and do it on more than just an individual server or six servers. They want an entire rack of servers or something that's meaningful for their solution. So what happens with some of the results? They buy an eco pod, they go, wow, we're really boobard, and they say, I have to get a home of five eco pods. What happens? I mean, what is typical thing that happens? In some cases, they'll give us feedback where we'll go retune other parts of our system for better efficiency. In other cases, you're absolutely right. They may have underscoped their power budget, and so now they can go and put more servers on the same amount of workload. Sometimes we can go tweak the clock frequency so they can change the quantity of servers that they're deploying. Sometimes they want to allocate some reserve for the future because they know they've got a growth strategy that they need to go fulfill. So there's a lot of great outcomes. I've heard some horror stories around CIOs who have made the bulk buys, big bulk buys, and they're used to doing all that stuff. So the power thing creeps up on them. Can you share any horror stories without mentioning names? Like just scenarios that caught people off guard from a power standpoint? Dropship pallets of servers, and it's like we can't install them, not enough power, have to go to another home of the building for an example. I mean, HP's actually got great technology around advanced power capping where we can really help our customers stay underneath those power levels. So I've got some great success stories where our technology's come to bear. Well, I mean, we're close to San Francisco. David Floer and I used to spend a lot of time with PG&E and, you know, a big data center in San Francisco can't get any more power. So PG&E had to implement this program where they basically would give credits to people for lowering their energy consumption so they had enough power to sell back. Well, and that's what's really great about our pods because we've been able to really tune the servers and the infrastructure right together in a closely coupled loop along with rapid deployment and rapid supply chain so that when customers get into a crisis with respect to power in the data center, they can go and deploy those very quickly. So do you expect people to... I mean, I've never seen, actually, people really lower the power bill. They just sort of, you know, keep it the same and then put more resources behind it. Is that the same scenario here, do you think? Well, with respect to the Redstone platform and what we're creating in the extreme low-energy processors and the products we're creating on that, we do expect them to lower the power bill and we've got awesome results that we've seen in the lab on the benchmarks that we've already run and we think that there's a lot of promise here and so that's why we're really opening this up to customers. All right, you heard it. Do you want to lower your power bill? Give these guys a call? Yep. You guys are going to be at... We started HP Discover, the pods were fantastic. Yep. Are you guys going to be in Vienna? Some of us will be there, yeah. We'll be in Vienna. I think that's in two weeks, perhaps somewhere around there. Okay, so besides setting up on the site, what are people going to expect from your program over the next year? What's your... Yeah, so over the next year, we expect to introduce more products into the lab and then also you'll see the partner ecosystem begin to build, so we'll get a more... a nice stable operating system and operating environment for customers and then we'll also grow to other regions across the world. So expansion to other facilities. Yeah. Anything out in the West Coast because HP Labs is here, it's not going to be part of HP Labs. We don't have plans for the West Coast right now but I wouldn't rule it out depending on how successful we are. Okay, Gerald Klein, great to have you on Program Director for Hyperscale Engineering. Thanks for coming on theCUBE. Thank you. Appreciate it. Good to see you. Appreciate it. Great.