 From Las Vegas, it's theCUBE. Covering InterConnect 2017, brought to you by IBM. Okay, welcome back everyone. We're live here in Las Vegas for IBM InterConnect 2017. SiliconANGLE is theCUBE's exclusive coverage of IBM InterConnect. I'm John Furrier, my co-host, Dave Vellante. Our next two guests, Eric Herzog, Vice President of Market for IBM Storage. Good to see you again, you were on yesterday, and Marco did, who's the manager of customer success and partners of Spark Cognition, a customer. Guys, welcome to theCUBE. Good to see you again. Welcome to the first time. Thank you. Thank you. Okay, so what's the, we're going to talk about some storage we did yesterday, but to get the customer here, what's the relationship, why are you guys here? So, we provide the storage platform. They use our flash technology. Spark is a professional software company. It's not a custom house. They are a software company. And Spark, not related to Spark, open source. Just to name Spark, Spark Cognition. Spark Cognition, yes. Make sure you get that out of the way. You go ahead, continue. So, they're a hot startup. They have a number of different use cases, including cybersecurity, real-time IoT, predictive analytics, and a whole bunch of other things that they do. And they, when the customer goes on-premise, because they deliver either through a service model or on-premise, when it's in their service model, they use our flash and our power servers. When it's on-premise, they recommend, here's the hardware you should use to optimize the software if the customer buys an on-premise version. But they offer it both ways. But part of the reason we thought it'd be interesting is they're a professional software company. A lot of the people here, our regular developers, in-house developers, in this case, these guys are a well-funded VC startup that delivers software to the end user base. So, tell us more about Spark Cognition. Give us the highlights. Yeah, appreciate it. So, Spark Cognition, we're a cognitive algorithms company. We do data science, machine learning, natural language processing, kind of the whole gambit there. We're working, we have three products. Spark Predict is our predictive analytics, predictive maintenance product. Spark Secure is our network log security product. And Deep Armor is a machine learning endpoint protection product. So, in that, you kind of hear, we're in the IoT, the industrial IoT, the IIoT of things, and also in cybersecurity. We've done use cases, and other machine learning use cases as well, but the predictive maintenance and the cybersecurity are two most advanced use cases, industrial areas. So, we've been around about three years. We have around 100 people. I appreciate Eric talking about how well-financed we are and how our success is really is budding thus far. And we're happy to be here. And where are you guys located? Now, we're based at Austin, Texas. Another Austin, yeah, Austin, Texas. It's dominant with Austin. So, it's good to have finance, and you can't go out of business if you don't run out of money. Talk about the industrial aspect of it. One of the things that is hot, it's not a mainstream conversation here, has blockchains, the big announcement, but IoT is a big one, but industrial IoT is interesting, because now you have the digitization of business as a big factor, and that data is going to be thrown off massive analog digital data now, so analog to digital. What's going on there? What are you guys doing there to help, and what does the storage fit in? Yeah, I appreciate that. So, IoT, industrial, it's obviously, there's big clients there. There's a lot of value in this information. So, for us, it's predictive maintenance is the big play. A study I read the other day by Boston Consulting Group talks about how it services and applications and the industrial internet of things. There's an $80 billion market in the next five years with predictive maintenance leading the way as the most mature application there. So, we're happy to be kind of riding on the front of that wave, really pushing the state of the art there. So, predictive maintenance is valuable to clients because the idea is to predict failures, do optimization of resources, so to get more energy out of your wind farm, get more gas out of the ground, you name it. So, having the software that can provide those solutions efficiently to clients without a lot of startup at each new iteration, so having a product that can deliver that intellectual property efficiently is important. And the whole goal is to be able to reduce maintenance costs and extend the useful life of assets. So, that's what SparkPredict is our product. SparkPredict, our product in SparkCognition has been laboring to do. We have a successful deployment of 1,100 turbines with Invenergy, which is the largest wind production company in the United States. We're doing work with Duke, Nextera, several other large electrical production companies, oil and gas companies as well. So, we're in Austin, we're near Houston, and we have a lot of energy production opportunity there. So, predictive maintenance for us is a big play. So, you guys did a session this week, you hosted a panel, is that right? And so, I mean, no offense, but what we're talking about now is really even more interesting than storage. But so, but it's a storage panel you were hosting, right? So, what was the conversation like? The conversation around that was we had three software companies, SparkCognition and two other software companies, and then we had a federal integrator. And all of them are doing cloud delivery. So, for example, one of the other software companies, MediCat delivers medical record keeping as a service to hospitals, okay? So, they're doing predictive analytics and predictive maintenance, and also some cybersecurity out. So, there was three professional software companies and integrator, and in each case, the issues were, A, we need to be up and going all the time, and the user doesn't know what storage we're using, but we can never fail because we're real time. In fact, one of the customers is the IRS. So, the federal integrator, the IRS cloud runs on IBM storage. The entire IRS runs on an IBM cloud, our storage, but it's their cloud. It's their private cloud that they put together that the integrator put together. So, the idea was when you've got a cloud deployment, there's two key things your storage has to do. A, it needs to be resilient as heck because these guys and the other two companies on the software side, if they can't serve it as a service, then no one's going to buy the software, right, because it's software as a service. So, for them, it's critical in their own infrastructure that it be resilient. And then the second thing that needs to be fast, you've got to meet the SLAs, right? So, we think of the systems integrator at the IRS. What do you think the SLAs are? And they've got like 14 petabytes of all-flash. You forgot they're cheap. You got resilient as heck, lightening fast. You've got to be dirt cheap too. Well, of course. They went all three, right? Now, so, you have this panel. So, Jenny, what were Jenny's three? Industrial-ready, cloud-based, and cognitive to the core, right? So, you guys are, I'm on your website. It's cognitive this, cognitive that. So, you're cognitive to the core. You're, presumably, you're using industrial-ready infrastructure and it's all cloud-based, right? Yeah. Talk about that a little bit and I got a follow-up. Yeah, so, you know, to tie in to what Eric is saying about the premium hardware, the cloud opportunity, you know, for us to be able to do AI software, to be able to do machine learning models, these are very intensive applications that require massive amounts of CPU, I.O., fast storage. To be able to get those, to get that, the value from that data quickly so that it's useful and actionable, takes that premium hardware. So, that's why we've done testing with Flash System with our cybersecurity product. You know, one of the most innovative things that we did, and, you know, in the previous year, was to move from a traditional architecture using X8664, we had a cluster of eight servers there, brought that down to one Flash System array and we're able to get up to 20 times the performance with doing things like analyzing, sorting, and ingesting data with our cybersecurity platform. So, in that regard, we were very much tied closely to the Flash System product. So, that was a very successful use case. And we actually, we offered a white paper on that, if anyone wants to read more, that's available on the IBM website. Where do you find that? Search it. Yeah, it's on ibm.com and it's basically how they used it to deliver software as a service. Where do I search? If you search Spark Cognition, IBM, you'll find it on Google. So, my other question, may follow up these, you talk about these IoT apps, which are distributed by their very nature. Can we talk about the data flow? What are you seeing in terms of where the data flows? Everybody wants to instrument the windmill. You got to connect it, then you instrument it. Where's the data going? You're doing analytics locally, you're sending data back. What are you seeing in the client base? Yeah, that's a great question. In those, in the field use cases for the wind turbines, for example, so most of our clients, they already have a data storage solution. Like, we're not a data storage provider. And the reason, someone asked me this yesterday in a different conversation, they said, why are wind turbines so much, so ripe for the picking? And it's because they're relatively modern assets. They were built with the sensors on board. They've been collecting the data for, since the invention of the modern wind turbine, they've been collecting this data. You know, it's generally, it's sent in from the field at 10 minute intervals, usually stored in some sort of a large data center. For our purposes though, we collect sort of a feed off that data of the important information, run our models, store a small data set. You know, a few months, whatever we think we need to train that machine learning model and to retrain and balance that model. So that's sort of an example where we're doing the analysis in a data center or in the cloud, sort of off out of the field. The other approach is sort of an edge analytics approach. You might have heard that term. That's usually for smaller devices where the value of the asset doesn't justify the infrastructure to relay that information and then deploy this large scale solution. So we actually are developing edge analytics solution, a version of our product as well. Working with a company called FlowServe, they're the world's largest pump manufacturer company to be able to say, how can we add some intelligence to these pumps that may operate near a pipeline or out in the oil field and be able to make those machines smarter even though they don't necessarily justify the robust IT infrastructure of a full wind turbine fleet. Is there a best practice that you guys see in terms of the storage? Because you bring out edge in the network. Great point. A lot of diversity at the edge now from industrial to people. But the data's got to be stored somewhere. I mean, is there a best practice, is there a pattern developing that you're seeing in terms of how people are approaching the data problem and applying the algorithms to it? Just talk to them. Move the data. Push the compute to the data. Thoughts on what you guys are seeing in terms of best practices. Well, so one of the other companies that was on the panel also is doing predictive modeling and they take 600 different feeds in real time and then munch it for mostly for industrial markets but mostly for the goods. So the raw goods that they need to make a machine or make a table or make the paper that is used behind us or make the lights that are used here, they look at all that commodities and then they feed it out to all these consumers but the companies that build these products. So for them, they need it real time. So these storage that's incredibly fast because what they're doing is they're putting on super powerful CPUs loaded with DRAM but you can only put so much DRAM in a server, right? And they're building these giant clusters to analyze all this data and then everything else is sitting on the flash and then they push that out to their customers. So slightly different model from what Spark Cognition does but a slightly simmer except they're taking in from 600 constant data sources in real time 24 by seven, 365 and then feeding it back out to these manufacturing companies that are looking to buy all these commodities. You have software defined in your title, right? It was the kind of big buzzword a few years ago and everybody kind of wanted to replicate the public cloud on prem. We think of it as programmable infrastructure, right? Set it and then you can start making API calls and set SLAs and thresholds, et cetera. Where are we at with software defined? Do you guys, does it resonate with you? Is it just an industry buzzword but I'll start with- So for us, we're the largest provider of software defined storage in the world. You know, hundreds and hundreds and hundreds of millions of dollars every year. We don't sell any infrastructure, we just sell the raw software and they use commodity infrastructure, whatever they want, hard drives, flash drives, CPUs, anything they buy from their local reseller and then create basically high performance arrays using that software. So you create on the own, everything is built around automation so we automatically can replicate data, snapshot data, migrate data around from box to box, move it from on premise to a cloud through what we call transparent cloud tiering. All of that in the software defined storage is done based on automation play. So the software defined storage allows them if you will build their own version of our flash system by just buying the raw software and buying flash from someone else, which is okay with us because the real values in the software obviously, as you know, so that allows them to then create infrastructure of their own but they've got the right kind of software, they're not home brewing the software, it's all built around automation. So that's what we're seeing in the software defined space across a number of different industries whether it be cloud providers, banks, we have all kinds of banks that use our software defined storage and don't buy the actual underlying storage from us, just the storage software. So do you, I mean, you may not have visibility in this but I'm getting kind of geeky on it but do you guys adopt that sort of software defined mentality in your approach or? Yeah, so for us, software defined storage is something that we've deployed for our proof of concept evaluations. The nature of the work that we do is that the solution is innovative to the point where everyone needs to have some sort of proof point for themselves before the company or the client will invest in a large scale. So software defined storage and sort of embracing that perspective has allowed us to deploy a small scale implementation without having our own dedicated hardware, for example, at different clients. So that's enabled us to spin up an instance quickly to provision that small scale deployment to be able to prove out results at a low cost to our clients. So that's where we've really leveraged that approach. We also have used a similar approach in the cloud where we've used multi-tenant environments to be able to support our cybersecurity product, Spark Secure, in a multi-tenant cloud hosted environment which brings down delivery costs as well, allows us to slice up that data and deliver it at a low cost. As far as our large scale physical deployments for the asset monitoring and such, we really, we generally end up with a piece of a flash system or a flash storage, bare metal deployment because that speed is critical, whether that's, you know, the client wants to have instant monitoring of a critical asset or they have a financial services use case where we're looking for anomalies or we're looking for threats on the cybersecurity landscape. You know, having that real-time model building and model result is very critical. And so having that bare metal flash system type installation is kind of our preferred route. The only other thing I would say on that is you asked earlier about sort of our approach. So for us, the security of data is very important. Most of our assets are what are called critical assets. And so clients are very sensitive to the security of the data. And so some are still uncomfortable with sort of a cloud deployment. So another reason why we have a affinity for the hardware deployment with IBM. So why IBM? So our company has really deep roots with IBM. My founder, Amir Hussein was actually on the board of directors of the original IBM Watson project and as well as Manoj Saxena was the original GM of the IBM Watson program. So we have just a long relationship with IBM. So we have a lot of mutual interest and respect for the entity. We've also found that the products are superior in many ways. You know, we are hardware agnostic and we're an independent advisor to our clients when it comes to how to deliver our solutions. But you know, our professional opinion, based on the testing that we've done is that IBM is a top tier option. And so we continue to prescribe that to our clients when they feel that's appropriate. They make that purchase through IBM. But yeah. Great, great testimonial. Eric, excited to hear that. Nice testimonial for you guys. Congratulations. He's done several panels with us. And again, part of the reason for here was A, all about IoT, which they're all into. All about cognitive, which they're all into. And to show that you can do a software as a service model, even in-house, you know, they happen to be a professional software company. But if you're a giant global enterprise, you may actually do software as a service to your remote branch offices, which is very similar to what these guys do to other companies. So this gives them an example. The other two software companies the same way to show in-house developers, if you're going to have a private cloud, not go public, you can deliver software as a service internally to your own company through, you know, the dev model and do it that way. Or you can use someone like Spark Cognition or Metacatter or the other companies that we showed, you know, Z-Power, all of which were using us to deliver their software as a service with IBM Flash technology. And you're using Watson or Watson Analytics? Yeah, so we have done integrations with Watson for our cybersecurity products. We've also done integrations with Watson, rank and retrieve using the NLP capabilities to advise the analysts, both in the predict space and in the secure space. So sort of an advisor to say, you know, what a client user could see something happening on a turbine and say, you know, what does this mean and using a Watson corpus? I was going to add one thing to, we were talking about why IBM, you know, IBM really has been a leader in the space of cognitive computing and they've invested in bringing, nurturing small companies and bringing up entrepreneurs in that space to build that out. And so we appreciate that. I think it's important to mention that. All right, Mark, thanks so much for joining and the great testimony, great insight, and good luck with your business. Congratulations on the success startup, taking names and taking butt. Eric, great to see you again. Thanks for the insight. And congratulations on the great, happy customers and see you again. Okay, we were watching theCUBE live here at Interconnect 2017. More great coverage. Stay with us. We'll be more after this short break.