 New York City, this is SiliconANGLE Wikibon's exclusive coverage of HP Moonshot here, special event where the server business is changing, the computer business is changing, everything is changing, everything is transforming, we're going to break it down for you. I'm John Furrier, the founder of SiliconANGLE.com, join with my co-host. I'm Dave Vellante of wikibon.org, and Patrick Morehead is in the house on theCUBE. Patrick's the president and principal analyst at More Insights and Strategy, a fast-growing consultancy. Patrick, welcome to theCUBE. Hey, thanks for having me on. Good to see you. You know, Mark Potter was saying in the Q&A session that you've been following this from the early days. You kind of had the sort of inside track on what's been going on. So now that you've seen this come to life, you know, what do you think? What can you share with us? It couldn't have all been pretty, but it looks like a nice package. Yeah, it's really impressive. And even going back five, seven years ago, been doing a lot of work in GPU compute with accelerators and things like that. The coolest thing is that bringing this to more of a mainstream market. Now, you know, this is a scale-out environment, which some people won't call mainstream, but in the context of scale-out, this is really cool technology that's delivering real value. So when you look at the changes, the C changes that you've seen over the last, you know, decades, let's say, where do you put this one? Well, I would put this right up there with the transition from proprietary UNIX servers to standard X86 servers. And the biggest driver here is the fundamental shift to mobility and the internet of things. And, you know, let's say you require for every 10 smartphones you need one server. We're shifting. Literally a billion smartphones, excuse me, a billion phones that were sold last year. And imagining that it was all transitioning to smartphones, and imagine that the rate that people are consuming that, and you add on top of that, that all of these different devices like from refrigerators to watches to Google Glass to even having intelligence and clothing are all going to be connected, today's data centers just can absolutely not take that. So what is it about Moonshot that makes it so well suited for those workloads and apps? Is it the ability to handle diversity? Is it the packaging? Is it the power consumption? Talk about that a little bit. The biggest change technologically is moving from this general purpose, very large core environment to literally having a server card or processor for the best type of application. So in the Moonshot mix, you're going to have CPUs, GPUs or graphics, digital signal processors. You're going to have FPGAs, which are used for just heavy duty number crunching. And you put that all in there, you can literally, and you heard it up on stage today from HP's customers, they're going to have specific servers for a specific application, whether it's genome mapping, whether it's cloud gaming, whether it's high performance computing, gosh, deep packet inspection is a great, great example of this. And that's a fundamental change. Now, the second major change here is that this isn't an HP thing only. OK, they've opened up and created an ecosystem where you heard about five partner vendors today, a year from now it could be 15, along with OS providers, OS tool providers, application providers that get into the game. And as we've seen from Apple and iTunes, in the App Store, when you create an ecosystem, people create stuff that you never would have even thought of. Obviously, the shift to this new modern era, as Dave and I always talk about on the Cube and what we talk about on SiliconANGLE and what he researched on Wikibon is this modern era we're moving into. And all those mobile devices is just one example, I mentioned Internet of Things, you've got telematics, booming, et cetera, watches that are going to have all kinds of sensors on them. You name it, and you wrote a nice piece on Forbes today outlining all that. But it puts pressure on the data centers and the infrastructure that powers that. Infrastructure is a service, platform is a service, and a lot of stuff is going on, OpenStack, AWS, all that stuff is happening. But I want to ask you about the traditional IT environment. Traditionally, scale up commercial software, buy some Oracle license, buy some SunBox or HP boxes, a lot of gear, a lot of bare metal, rack and stack. To get stuff going, it takes a lot of money and effort. That's why cloud has been so great, it's why Amazon is doing so good, that's where it was going there. But I want you to talk about the challenges of the mindset of that IT enterprise that has to transform over to a world that is dominated by developers now, software, and scale out where you need commodity hardware or industries data hardware as HP wants to call it. But ultimately, it's a lot of gear, a lot of servers, a lot of storage spread all around the infrastructure, utilized maybe not 100%, power is huge issue. These are all cutting edge trends. How does that modern era relate to this moonshot, and where do you see it going? Yeah, so to understand kind of where we're going, we got to know where we are today, which you outlined great. But where do we come from? That really defines the mindset of current IT. Enterprise IT has been there when there were mini computers and main frame. There are still main frames out there, obviously. But where you had a different box for different applications. I cut my teeth with NCR with financial applications where you had a system 360 for the general ledger. You had tandem for ATMs. You had deck for treasury. And it ended up being a mess. And then we went through this homogenous centralization with virtualization and pulled a lot of that stuff in with client server to where we are today. So there's going to be a natural inclination to push against this as where it wants to go. But you mean the scale out? Scale out, yeah. Push against the scale out. People are going to pay and let you do it. The biggest objection to scale out is security. But in three years, three to five years, people are actually going to change it to the cloud is the most secure place to be in a scale out environment because the cost to do that is a standard enterprise won't be able to afford it anymore. So the way this is going to come in is there will be a corner of the data center where somebody can get this absolutely immense speed up or they can pull power down to a minimum and they'll cut their teeth. Let's call it even a major bank where they're one of the largest web servers. They can put Apache in as an example. And doing a flat file Apache serving could be a great place to start. Hybrid clouds is another one where you can take advantage of the scale out technology without actually having to put it on your premises. So there's another component to this too Patrick that I'd like to talk about and scale out and hyperscale specifically is the hyperscale customers will spend a lot of engineering time trying to like hyper automate such that they don't ever have to have a human get involved in doing things like replacing failed components. They'll just let the thing die and then throw it in the woodchip when it's done. That's a hard engineering problem and a lot of companies don't have the resources to do that. So what do you see as the software infrastructure that has to evolve in order for that to become a reality in the traditional enterprise. Great question. I mean currently today the scale of data centers they roll their own OS literally down to the kernel. They buy nothing prepackage they do with themselves in addition to the management services that they have which sometimes is just what failed so I can go throw it away because it has automatic fail over. Okay and they're using open open stacks like the lamp stack to be able to do this open flow for networking and even you know literally open stacks. So they have the ability to do that over time as the economic opportunity grows I see people like Microsoft getting into the game more heavily supporting this heterogeneous computing environment because it does take a lot of work to make your OS work with CPUs, GPUs, DSPs and FPGAs but if those packaged operating system providers want to play in this space they're going to have to support this. So what do you make of Facebook's open compute initiative? There's some similarities in the multi-personality capability here. Of course Facebook is not a server vendor but they're defining trying to find a standard that server vendors and other ecosystem partners can actually write to. Do you see Moonshot as competitive to OCP, complementary? How do you see that all shaking out? So I will I see Moonshot as very comprehensive and wide in the fact that it supports multiple compute engines with a hardware architecture that you can just slot different things in and out. I see OCP as a narrower opportunity here literally redefining an architecture every revision that they have literally down to changing the width of the rack. So I don't think that there's one architecture that's going to win them all because this opportunity is so big but I think it's going to be a challenge for open compute to change everything down to the bare metal every time that they want to make a revision and I don't see open compute supporting accelerators like GPU compute, DSPs or FPGAs which are an essential part of this heterogeneous computing. Okay so OCP is narrower, Moonshot is going to find its way into more use cases and who knows potentially even Facebook. I mean could you see a Moonshot going into applications like that? I mean why not? Well I could in Facebook. I mean Facebook isn't they'll take the best technology in and I could see an HP Moonshot system inside of Facebook for specific usage models that open compute just doesn't address today. The hyperscale guys are also they kind of don't like frills. Is Moonshot defriiled enough to penetrate that market in a big way and or do the hyperscale guys actually need some of that? Are they maturing those two worlds coming together? Yeah so most scale up guys don't need the management chips and management software that comes with it. They're just not going to use that and I think it really is going to come down to an element of pricing in one instance they also are very comfortable with moving vendors in and out at a very rapid pace so we'll wait and see how that rolls out. I do think though the flexibility to be able to take these cards inside and out and having a common chassis there's a lot of value to that. So instead of rolling in rolling out a complete box you can literally change up cards on the fly if you want to address capacity issues in your scale up data center. Patrick HP Moonshot uses partner silicon partner software. What does HP bring to the table outside of the packaging that you see in this deal? Well the biggest thing is the ecosystem in getting people unwrapped with new cartridges, OSs that are compatible tools that work with the OSs to make sure it's quote unquote certified. Ironically for some government applications HP will never see the cartridge in that they're doing something very special with it that quite frankly they don't want other people to know about. So they're fostering this ecosystem of hardware and software to get it to this point where honestly you don't know where it's going to go when you've got an ecosystem going like this. The key, the challenge is going to be that some midterm profitability for the partners, right? I mean nobody would be at the Apple App Store if they weren't making money and didn't see an opportunity. That's going to be the challenge having a lot of partners in with a very high ramp up to keep people's mind in the game. Because it, while it's easier to develop a cartridge than as a server, it still takes a decent amount of resources. So I'll get your take on the developer community because obviously you mentioned the old days, IBM ran the general ledger NCR, then they put it all together. In today's market you see similar challenges where you have that general purpose model going away where you have custom builds and people saying, yeah, once they build their own open compute summit, which might not make it in time, might be around. But ultimately people want vertically oriented applications and infrastructures vertically integrated to that where open source could play a role. So in this custom game, how do you see that playing out where people are saying you can still get high performance with custom? Is that compatible with the scale out and how would you talk to people about that challenge and that question? So the reason why the main frame and the many ran into challenges is that the rate of acceleration was nowhere even near the speed as general purpose X86 servers. And the difference here is the reason why this has a much better chance of offering the speed and the flexibility is because you have this ecosystem of folks. A lot of people who are in the smartphone business who are churning out new technologies every three to six, 36 months. And with a stable lack of better term backplane, popping these upgraded cartridges in place when you need them, I think narrows that or alleviates that challenge that I see. Patrick, what's your, you've been in the semiconductor business as well. You've got a varied background. What's your prognosis for Intel? You're seeing just the sheer volume of low power chips like arm or just dwarfing traditional X86 and volume wins generally in the semiconductor world. So what do you see there? Can Intel move fast enough to low power? Is it fundamentally harder for a company that's got such fixed assets to do that? What's your take? The irony of it is that there's nothing about X86 that makes it higher power. Okay, and I think Intel did a good job with their smartphone processor called Medfield to show that versus arm, they are as efficient. To the data center, it's all about the rack, right? Power at the rack or in the fleet of racks. Intel is perfectly capable of doing that. It really comes down to a business model decision. Now interestingly enough, if you're packing 1,800 processors in a server, and you compare that to the economics of what it would be to a Xeon, and you own the fabric, there's not as much difference as you might think in it. I think what Intel needs to focus on is those other accelerators that go in there. They currently have Xeon Phi Accelerator, which is similar to, I would say architecturally to a X86 flavor with more of a GPU architecture. But they need to be considering other types of accelerators, too. Otherwise, that silicon revenue will go to somebody else. Yeah, I should actually ask the HP this, but you may know. I presume I'll be able to slot i5s and i7s into Moonshot. Yeah, I believe on stage I heard Xeons as well as well as Adams in. And there are a lot of workloads that fundamentally run better on big cores. So I don't think this is an either or, it's an and. Well, so I mean, so I tweeted out earlier today that this ultimately I think this. We looked at all the programs out there and right now the processors, the memory's too small, but to the extent that I mean, this seems to me to be the future of packaging. I mean, you wrote an article on Forbes saying goodbye vanilla servers. Yeah, I mean, do you buy that? I mean, is this the future of server packaging? This is absolutely the future fundamentally because we're running out of space and power and the only way to attack that is to grow the data centers bigger and be more prescriptive on a specific application for a specific server. Otherwise, the efficiencies don't make sense. So this is, and this will not only eat into blades, this will also eat into racks. And if you think about a small business in three to five years, are small businesses really going to be able to afford bringing in their own infrastructure and sticking in their closet? Or let's say if it's $0.10 on the dollar going to an Amazon or an HP cloud to do their open stack based applications in terms of HP. And with Amazon, they're proprietary AWS. So it's heat density and it's non differentiated heavy lifting. So we have two minutes left. So we want to get just a couple of quick questions before we do kind of your final take on. I want to ask you about the competition real quick because that hasn't come up yet. Because you're an analyst, you want to get your perspective on that. Got a comment on Twitter though that, sorry, the software defined server made me throw up on my mouth a little. And I guess he's referring, I just tweeted back, software lead may be a better term that we use. But I think people are missing the point about what this means. I don't think they're being as, it's all semantics. It's not like software defined as in the networking which open flow drove, but more specifically, a software enabled environment. Can you just give your view on that and then talk about the competition to Moonshot? Sure. So first of all, it is confusing, SDS and SDN, it is confusing. And it tripped me up a little bit at first. But essentially it means it all starts with the workload and the application. And you build out your server based on what you need. Do you need big compute, small compute? Do you need a DSP or an FPGA? And so it's literally based on the software. The server is based on the software. And I do think, as I said before, that is the future. Software and it enables software developers. So I think this is where the hype kind of dissecting with the hype is, OK, is HP just using the hype term to kind of mark it and put a stake in the ground. But specifically, it's software in the system, but it's also enabling software developers to write code. It is. It's also about a way of thinking about it, that it starts with the software. It doesn't start with the hardware, it starts with the software, and what you actually want to do with it, which defines how you configure your server in a moonshot environment. OK, final word, we're getting the hook here because we're on a tight schedule. What does this mean to the industry real quick, sound bite here? No, I think this is a huge day for the industry in that I really do think we're going to look back and say, hey, this was the beginning of something big, something which took us to the next level. Let's see if the ecosystem can step up and rally around HP. Obviously, we're big fans of this direction. We covered moonshot last year. This is the Cube special presentation in New York City for covering HP's moonshot. A lot more action stay right here. We've got a lot more interviews and we're going to get the signal from the noise here inside the Cube. I'm John Furrier with Dave Vellante. We'll be right back after the short break.