 Okay, welcome back, everyone. We're here live in San Francisco. This is SiliconANGLE's theCUBE, our flagship program. We go out to the events, expect to see them from the noise. I'm John Furrier, the founder of SiliconANGLE. Join my coach, Jeff Frick, general manager of our CUBE operations. Filling in for Dave a lot of things out on the road. Dave's watching us mobile app. I just talked to him on the phone. Dave, shout out to you. Get on crowd chat, tweets and stuff. We'll be watching. My next guest is Jason Snow, CEO of Cyclocomputing, CUBE alumni. Became famous literally overnight, internet famous on the CUBE Amazon, a reinvent conference. Our first time we've got the CUBE to one of the big Amazon shows and now we're never going to leave. We're like a tick. We're embedded into the system. We're not leaving. Jason, welcome back. Thanks a lot for having me, John. It's good to be here. Your last interview was one really was awesome. I still, it's one of my favorite interviews. Very memorable. A lot of passion, a lot of energy. And it was really our first exposure to Amazon doing a live show at the Amazon event. We've been customers of Amazon. And we're drinking the Kool-Aid, using it. Happy customer. But the question was, is the real deal? There's always what everyone's asking. And your interview was really awesome. So you're doing some real stuff right now on Amazon. Let's recap what's going on. Sure. And what's happening here at the show. Yeah, so we've definitely had a real uptick in usage since the 156,000 core run that we did in November. But we've also had a lot of great case studies where real customers have kind of come forward and talked about various aspects of life sciences and manufacturing or insurance and financial services have been doing work. So we had Johnson and Johnson talking about different storage and archival models at a conference recently, the Aerospace Corporation. These guys do some really cool rocket design for the Apollo missions back in the 60s. But most recently they did the rocket that brought curiosity to Mars. Now their future generations of rockets will in part be designed and simulated on AWS and Cyclo software. They're basically doing all of the engineering aspect of kind of literal rocket science. So there's a bunch of different use cases in large companies and government. But I loved it. Yeah, I mean normally we get guys on here like, hey, I'm the CEO of a company. I built an app on a crowd chat app on Amazon like us and blah, blah, blah. You're doing some pretty serious scientific computing, really kind of a little high powered stuff. So that must get a lot of attention. So like, there's a lot of guys going out there doing internet of things. You're hearing about Jeff and I were on an event with the GE and turbines, oil exploration, hospitals. Big use cases where computing and having huge resource available is out there. So that being said, that's pretty much done to you. We see the internet of things coming. How are you seeing that trend? I mean, it's obviously early on now, but what are some of the things people are doing on the high performing side? Crunching numbers, is it exploration, science research, what do you see? There's a large amount of use cases across life sciences. So it's everything from genome analysis to proteomics and drug design all the way through designing your clinical trials so that you take fewer samples from your patients. And when you have like pediatric cancer kind of clinical trials, you wanna be drawing blood from them as little as possible. So these kinds of use cases are definitely common. AWS has been a really enabling platform from an infrastructure perspective, but we have a large number of Fortune 500s as well that are kind of doing just your regular product design work. Steve Philpott, the CIO of HGST, which is a Western digital company. They make most of the hard drives that are in your laptops and your computers and even in S3. That company is now designing its next gen hard disks on their own hard disks inside of AWS. So it's kind of a skynetty thing, but... Steve's a very... The compiler for the compiler. Exactly. I don't get that. Hard drive for the hard drive. It's recursive technical computing, but in the meantime... Virtualization has caused some interesting things, right? You can just do some virtualization, you got now the cloud. This brings up the thinking different mindset. You know, the classic Apple commercial, think differently. What have you seen out there that's different, that's vectoring into mainstream, that's going to be disruptive? What's that foreign object that people haven't figured out yet that's coming into the market that's going to be the new disruptor? So I think the exciting thing about cloud, and I've said this before, maybe even on this show, is that you can now do things that you couldn't do before. When you got a car, you can travel more than 50 miles in a day without killing a horse, right? So nowadays, we can spin up a compute environment that's massive, if only for a few hours, to make it so that large Fortune 500s, Fortune 100s, can do amazing work. We've got people that used to only have a couple of 1,000 cores or a few 100 cores in-house, can now dip in, grab 8,000 cores, run several weeks worth of work in a day or two, and be able to return back a result. That's huge for risk management. It's huge for product design. Anyone that's designing a physical product is doing competition fluid dynamics. They're doing heat transfer simulation. All of these kind of core, heavy engineering applications can now be done much, much faster on a utility basis in the cloud than they could ever have been done before. So the last few things we talked to you guys about was actually around 156,000 core environment that was essentially, it would have cost $68 million if you bought it, but instead we used it for $18 and paid $33,000. And this was for a material science use case. Since then, we've actually had companies that are doing paints or other forms of materials design actually come out and say, look, we need to do this exact same thing to generate, next generate materials for manufacturing. So this is a really exciting time because the use cases are kind of exploding. It used to be genomics and maybe some forms of financial services, but now we're seeing things in manufacturing and energy in all kinds of different parts of the kind of the scientific and engineering computing spectrum. It's been very exciting. Well, it was great because Andy and his keynote talked about kind of a, I don't know if it's a re-going back to, but the supporting of an experimental kind of way of doing things. And then you're doing that on like super steroids. Yeah, that's good fun. Experimenting in ways that they can never even imagine was possible. I mean, are people kind of catching up to it? Do they sit there in their eyes roll in the back of their head as it slowly starts to sink in to really think about things that they couldn't do. They would never even think of potentially doing. Now that suddenly you've been able by bringing all this power to bear for limited periods of time to accomplish specific tasks. Yeah, so our user base has a majority of people running 40 to 4,000 cores on a day to day basis, day in, day out on a regular basis. But every time we do these larger scales and it started, we did a 10,000 core cluster in 2011 with Genentech. And then from there did a 30,000 and a 50,000 core. Last year we did a 10,000 server in that really large 150,000 core environment. Every time we do those, we get customers that are now like, hey, you know, I'll give you an example. In our pipe right now, we have a pilot with a very large financial services institution. You're probably going to need 20 to 25,000 cores to take a 20 day process and turn it into a one day process. So when we do that kind of really large run people come out of the woodwork and it's like, oh, well, 25,000 cores is no big deal. It's not 150,000, right? I mean, it seems reasonable now, even though if you were to think about it on a practical level, 25,000 cores a year ago was just astronomically large. But yet Amazon has the capacity that Cycle can go in and build the live computing environment for essentially a Fortune 100 company that needs to be able to run a real business process and get the results back as quick as possible. So how much of your business is just building and putting the infrastructure in play versus getting people to start to think in terms of those scales, that type of scale? Yeah, so we do the large use cases to help them with that, but I think a majority of what we've been helping with over the last year or two has been helping get ISV's applications so that you can push a button and have a computational fluid dynamics environment or a heat transfer environment or a genomics analysis environment. And you just use our software, you fill out a small web form and you have a compute cluster that does what the bare metal that you have in a house did, but it doesn't take you weeks to months to get it. It doesn't sit there for three years and depreciate. It has a much better cost profile and agility profile. Getting those applications and making it so that ISV's can very easily get onto cloud is actually critically important for the users because people don't want 156,000 core clusters. They want the solar material that's going to make a better panel and give them a competitive advantage. That's what they care about. So this is basically what we've been really focused on quite a lot. Jason, I want to ask you some questions about the market. Obviously, given your perspective, you're under the hood, you're doing some amazing things, huge cores, you're spinning up, but now you have three categories. The old guys being disrupted. Yep. The new guys been around for like less than 10 years and five years. The new disruptors being disrupted and then the new entrants. So you're seeing kind of like three phases of actors out there. The old guys, the IBMs of the world, HPs, oracles. The new guys, maybe a startup that got funded five years ago, maybe built a cloud of a certain direction that's an unwind pivot or whatever, but big funding. And then the new guys. Yep. What's the disruption? Those are the guys being disrupted on the far end. The new guys disrupted completely. And then this guy here is trying to figure out whether he's disrupting or being disrupted. Right. So either way, there's a lot of shifting going on. You have IBM, Google, HP, Oracle, EMC, Biblital, VMware, Microsoft, all coming out of the cloud. It's just go announcing a billion dollar cloud. I mean, come on. Yep. So I think. I mean, you started going, okay, we're in a cloud war. But the reality is it comes back down to what the reality in the market place is. Yep. What customers want, what are the technical solutions. So describe who's being disrupted? Who's the disruptors? And what are the key things for success in the future? In the future's cloud for ubiquitous compute, storage, application of one internet of things. What are those key levers? What are the pressure points? Yeah. So from a motivation perspective, talking about that first, right? And then the party's second. So in terms of motivators, the big one is access, right? So people that couldn't get something before, whether it's a big data environment or an HVC environment, but more specifically, it's a utility access to a computational application. That is now very quick. It used to be that your vendor had to say, oh well I know a bunch of guys over at HP they can ship you a few servers, we'll install the software on it, you could evaluate this for a while, and then maybe buy it. Now I can push a button and get a cluster that runs it in 15 minutes. Yeah, and they're hiring not interviewing people either. Right, exactly. So you don't have the administration coefficient, a lot of the internal guys that used to do the administration can now actually help users use computational science to better the business, or to use data analytics to better the business. So it's a much better focus of effort and focus on core competency. So the utility access is a big one, the focus is a big one, and then obviously the pay-as-you-go aspect is huge. So having, you know, amortizing the risk of utilization, that was the problem. When you used to buy a lot of infrastructure for computation, you'd have to do peak versus median usage analysis and figure out, all right, well we can afford to spend this much, but then you're jailed by that capacity because that's the biggest question you can ask going forward. Now that's not the case. That's foundationally, instead of foundation, you're kind of stuck with the footprint. Exactly, for three years or five, depending on how long your depreciation schedule is. Versus an elastic, no pun intended, or pun intended, more of an elastic model. So basically where Cycle comes in and what we've been really focused on is basically helping the ISVs to disrupt user access patterns without having to necessarily change the license model. So if I'm, you know, Western Digital as an example of the HGST, the subsidiary that I talked to reinvent, Steven Philpott's doing a great job of transforming that business, making the engineering more agile, making it more reactive to the market conditions. And that process is one of benefiting from elastic capacity, but it's also a person transition. So the other thing that's going on is kind of what you asked about, are we doing a majority of our time evangelizing? I think the market kind of comes around and that's why those really large use cases are critical, because it shows what the potential is. And then someone says, oh, well, you know, I never would have asked for 25,000 cores before, but now I'm going to do a pilot with one because I know it'll change my business. And that's a big, big difference. So those are the kinds of disruptions that are happening. From who it's going to disrupt, I think it's going to disrupt everyone. If you ignore elastic capacity, utility access to applications, you're not going to be in business forever. It's just the bottom line of it from a practical standpoint, whether you're a Dell or an HP or an IBM, or you're on the startup front, you should be thinking about what's best for the user. And this is one of the things I know you and I talked about before being in Bootstrap. You're myopically focused on the customer. And that's the same way Amazon is. So we basically have a key initiatives around making sure we're hitting customer use cases only. We don't build science projects. We build stuff that does real work for a user. And I think that applies to all three of those categories. So the traditional players, get myopically focused on the user and what they want. How do they want to pay for it? How do they want to be able to use it? What's the usage pattern? For the newer entrants, you probably have other innovations that you bring into the table, but at least bring it to the table in the right way around access as well. It's an absolutely critical function. And that's what it is. Expose that service to people. Exactly. And this is where we've been helping the application vendors a lot. I think there's been a lot of... So about developers. So obviously the big guys will focus on whatever they want to do. Their strategies, Cisco, IBM, HP, they all got big accounts. They have big install base. They have customers to listen to. The good news about IBM is they've got big customers. The question I have for these run developers, this seems to be a land grab for developers right now. Everyone wants to win the developer community. Every company I talk to, HP, IBM, I'll be in the developer ecosystem. Do you? Yep. Maybe you had one that needs a little bit old. Microsoft, some say Microsoft's developer community really hasn't been upgraded and modernized in some debate that. But the point is, the guys who's developing value, they're going to get inundated with spam. Work on my platform. So, okay, what's your advice to folks who have that strategy? How does IBM, how does Google, how do they win the developers over? How do they... In a way that's not in their face. Right. Yeah, I think the key thing is to make all of the APIs open and available and have really great partner programs where you're essentially subsidizing usage of your platform for those users so that they're able to develop for you. I know we've spent a significant amount of effort in supporting AWS and I know the AWS guys have done a great job of partnering with us around making it so we can grab a $68 million environment for 1500 bucks an hour. That kind of outreach is definitely an important aspect of things. From a cycle perspective, I can give you a first hand account on this though is where we interact with the applications because a lot of the engineering and scientific apps aren't built for cloud. The code base is not really oriented around that. It's oriented around infrastructure. So, one of the things that we've been doing a lot of is basically making so you don't have to modify your code. Basically, we make cloud infrastructure look like the HBC clusters that these applications run in already. That way, no developer effort is really required to make it run inside of the cloud environment. So, the other point I would make is that to the extent that AWS and all of the other parties you're mentioning can make it very easy to not really have to do any coding necessarily to your API to have your application deployable. That's a huge benefit, absolutely huge benefit. Let's talk about the game-changing nature of HPC. High-performance computing. Boy, that world's changed, right? Box-centric, data center-centric clusters. You mentioned HPC clusters. How is HPC changing and how does big data affect that? And is it going to be two camps, old and new? Are they blending together? I mean, you're doing things pretty creatively. You're looking at spot pricing and all cluster analysis. You can spin up stuff pretty quickly. How is that market changing? How is HPC changing? That's a great question. So, there's two things about that. First off is there are multiple different workloads that are actually called, they're called commonly HPC, but they aren't all necessarily HPC. So, there's a lot of folks that are running massively parallel weather work that is true HPC. It's very interconnect sensitive, et cetera. There's those class of workloads. There's also the things that are more high-throughput oriented and they're essentially geared at enabling people to find needles in haystacks or run Monte Carlo simulations, do business analytics, do genomics. Those are the high-throughput oriented workloads. And then there's the big data stuff where you have a really tight coupling with the data and the compute side of it. Those three workloads really have different properties. And what we're seeing is, A, the low end or the everyday use cases, the engineering your physical product side of traditional HPC, the cost performance is now, and not having the overhead of managing the infrastructure is now beneficial enough that manufacturing workloads, which normally would only be done in-house, are now moving kind of cloud. On the opposite end of the spectrum data, there's a big data side. There's a lot of usage of big data in cloud from the beginning. They kind of almost grew up together. They're brother and sister rather than first cousins. So in the middle, the high-throughput side, you can get tremendous time compression. And this is that 264 years of work in 18 hours compression that we got in November, right? Where we did a quarter of a compute millennium of material science in less than a day because we grabbed a very large amount of infrastructure. And that one isn't as closely related as the big data side, but it is a ripe use case. So my prediction, I guess, would be big data will continue to work in cloud-friendly manner because traditional HPC clusters shouldn't be doing big data problems. So using them in the cloud makes sense. High-throughput workloads pretty universally should be looked at in a cloud context. And the MPI workloads, there'll always be a section where Uber performance is important, but the everyday use cases will start coming to cloud as price performance kind of makes sense. And as Amazon keeps dropping prices, it's going to keep enabling more and more classes of science that didn't make sense to make sense on cloud. So I think it's an exciting time, obviously, if you're an engineer, a client, a life science researcher, all of those guys are going to tremendously benefit from these changes. How about the people supplying HPC solutions? You've got a lot of vendors who have old models out there. How are they going to adjust? You've got Intel makes components. You've got the boss guys making gear. So Intel, I think, is doing a great job. They're actually enabling through having Ivy Bridge available on AWS, like right out of the chute through having Sandy Bridge was actually available on AWS before it was available anywhere else. So that kind of close partnership definitely benefits the end user, but it also makes those processor types much more relevant to end users, allows them to get better compute per dollar than they normally would get. So you see Intel as a player in this? Oh, they're doing a great job, yeah. I definitely think there's going to be a lot of stuff we'll see out of Intel that'll be very interesting. I also think on the traditional box vendor that people selling, IBM selling its server business is not coincidental. So it'll be interesting to see how that market evolves, but the fact that you have one of the folks kind of getting out of that space is somewhat of an indicator. Where it'll land is. Well, HP, IBM, you're seeing Oracle getting into the engineering systems, kind of purpose-built, high-end, high-performance, Godboxes, I guess they used to call them Godboxes. Yeah, that's right. Big iron. So Jason, you mastered the art of spinning up lots of compute. What about the data side, the massive amounts of data? Clearly when you got to move that stuff close to the computer, move the computer close to the data, it's real stuff that's got to go. That's right, yeah. What's happening there that's gonna enable you to apply this massive compute scale on similar kind of scales of the data? So I think there's gonna be a lot of demand for bandwidth, no doubt, over the next decade. We see the data volumes going up faster than my cable connection to the internet is, so that's gonna be a big factor. We've been working on a product called Dataman that basically is a scheduler for data movement. So it's not a protocol, it's actually protocol agnostic, so it'll work with native Amazon APIs. It'll also work with things like R-Sync and we're working on Grid FTP and other protocols for moving the data. But the timing and scheduling and workflow pieces are critically important and we actually have a lot of interesting IP and a lot of interesting software in those areas. We did a 1.21 petaflop computer with 150,000 core. We now have a 1.2 petabyte data scheduling which is taking in encrypting genomes and moving them into Glacier. So they can be archived on an exceptionally cost-effective basis. And we think there's gonna be a massive market around the workflow pieces because there are a lot of hard problems in that space and no good solutions. No solutions, period. Much less good solutions. You pump it into physics too, right? I mean you move it, you move bits around. Well and not only that, you have audit and reporting and which user needed which data, transferred where, at what time and those kind of compliance and charge back and reporting use cases. There's really not a good tool there. Dataman's really kind of the only thing in that space that does a phenomenal job on all of those. And there are a lot of great tools for cloud connectivity. So there's various vendors that have appliances and other things that connect to cloud but determining when, where and how in the workflow piece of putting the data into those environments is still a really untapped problem and that's something we're aiming right at because essentially a lot of the stuff we have to move data around in a 156,000 core environment, turns out to be really relevant in moving data off of a scientific instrument onto the filer, over to a research environment that's running compute against it. Those kinds of movement use cases and tracking all of that is actually critically important. So it's the new area around scheduling for us is going to be in the area of data scheduling, data transfer. Yeah, especially the Internet of Things that volume is only going to go up, right? Right and tracking requirements are going to get that much bigger because what happens if you essentially drop a message or miss things about a live feed of telemetry data from cars or crops or what have you. I mean, there's a lot of very interesting audit use cases there that I think are going to be exciting over time. So what's next? What's the next hill to climb? So, the next ones for us are going to be, again, talking about making exceptionally easy to spin up an entire research environment with a push of a button with all the applications you need. You'll see some use cases coming out about us. We actually announced today that the Scientific Software Schrodinger Material Science Group, this is Matt Halls. He's a stellar researcher and basically runs Schrodinger Material Science for the product side. We now can offer that to customers. So what we did in that 156,000 core use case is no longer a one-off. You can actually push a button and be able to get a very large cluster and that's a commercial offering with them. So we do that in concert with them. We're really excited about it. You'll see more and more of those kinds of packaging of different applications in different sciences and engineering and quantitative finance areas. The other thing that I think is going to be really exciting is the data movement stuff that you just touched on. I think basically being able to make it really easy to manage big data, not from a compute perspective and analysis, tools like Hadoop and Cassandra and whatnot work really well, but what if I want to take my data out of Hadoop and run it directly into Glacier and archive it uniquely? Those kinds of use cases are untapped and things that I think we can do an excellent job of enabling. So that's the other side of it. And then lastly, we're going to continue to do utility supercomputing every chance we get where we can talk about it. So we have some use cases that are tens of thousands of processors right now that we don't have customer permission to talk about. But as we get larger and larger use cases in that space, we'll come to you guys and tell you about some really cool science and how somebody's changing the world in their area but by utility access through the cloud. It's what AWS enables is awesome, especially when you put it together with our software. It's been very exciting. Mind bending, I think, even. Okay, Jason, thanks for coming back on theCUBE. We're great to have you, great insight. Always great to sit down and extract the data out of your head and share with the audience. Folks, this is theCUBE. It's what we do. We go out to the advanced instructor center from noise. Jason Stowe, Guru, CEO, something like doing amazing stuff with the cloud. Creativity is being unleashed. Productivity, game changing, disruption, and innovation. This is theCUBE. We're right back live in San Francisco for Amazon Web Services Summit after this short break.