 Live from San Francisco, it's theCUBE. Covering IBM Think 2019, brought to you by IBM. Welcome back to Moscone, everybody. The new improved Shiny Moscone Center. I'm Dave Vellante with Stu Miniman. This is day one of IBM Think, and you're watching theCUBE, the leader in live tech coverage. Ed Walsh is here. He's the general manager of IBM Storage and Software Defined. Ed, great to see you again. Always a pleasure. Thanks very much. I love the venue, gets us out of Las Vegas. No offense to our friends in Vegas, but the Shiny, it's been a long time coming, Moscone, but it looks really good. I agree. 30,000 people expected, so you must be excited. No, I think we have a lot of things that you're going to see announcements. I also, I think you're going to see some refinement of the overall message. I think it's going to be exciting week. So it's kind of, I'm talking to you before all the keynotes, but it's an interesting week for sure. So tease a little bit. What can you tell us? Okay, so my background has been outside IBM, coming in to run storage, but I've heard, I've said it a couple of times, my strategy is to drive the overall storage, but also get more aligned with what we're trying to do overall IBM, because that's the strength of IBM, right? Really helping the clients move forward and infrastructure matters. And what you're seeing is, I think the market's coming in our direction at IBM. And I'll give you a couple of things. You're going to hear a lot about hybrid multi-clouding and say AI at scale, right? And you can see that messaging, but that's where the market's coming at us. You saw Red Hat talk about it, all of our competitors are now doing that as well. But when it comes to, once you start about hybrid multi-clouding, we think it's like, we're going to talk about the chapter two, right? Chapter one was the first 20% of workloads and it was all about these application-driven events and thing about Office 365, et cetera. But the 80% of the workloads are still on-premises and there's a reason they're on-premises, but what now is like the people that next phase are going to be leading by the organizations that have the mission critical data and how to do that. And there's a role for hybrid which plays perfectly for us. And can you help connect the dots for us? Because we've watched the software-defined wave come through storage, but still, many people on the outside, they'd be like, okay, well, storage is a bunch of boxes sitting in my data center and all my new apps are being built here or I'm sassifying things there and it feels like death by a thousand cuts to the traditional storage market. So help explain kind of what the modern storage market is. Data's at the center of everything so we know that that's a huge thing that's a good tailwind for storage, but help bring us inside your business. Yeah, so I think in general, everyone's trying to be data-driven and it's easy to say hard to do. And everyone's, the platform you're going to do that is a hybrid multi-cloud and hybrid being, the reason we're using that terminology as an industry, but it's also, there is a role for on-premises and how do you easily connect, but it's getting the same agility and performance and cost benefit on-prem and then extend at the right time for the cloud. So where you see us, we're looking, we focus on, we deal with a lot of different clients. Some are advanced and some, I would say, more laggards. And so we share a lot of stories where people are being successful being data-driven. And we see it fall into kind of three, what I'll say is we try to success criteria or areas, right? So one is people just modernizing, going from traditional to go private clouds, making sure they extend the use of benefits of public cloud and be more data-driven. So some people are spending money to save money and some people are spending money to make money. But even in traditional environments, you see the CIO having a bigger voice and we need to push them and show them how to do that, right? We also see another section of people really driving AI in general. And we see that in definitely a hybrid multi-cloud environment where we see a large on-premises, but you're always looking for the different data sets that are going to be in the multi-cloud that they need to bring it together. And we see that being a very interesting, and in fact, how do you start with someone doing, we believe we have the best storage for AI, but we can scale from the smallest to the biggest AI supercomputers in the world, right? Driving 1.5 terabytes a second. They're huge monsters, but you can start small. But what we find is people are just getting started and we have to meet clients where they are. So part of it in these different areas is meet them where they are and there are different parts of journey on AI and it's having them get going. But then really, how do you scale? You're going to hear a lot about this. How do you scale AI? Everyone has these random acts of AI or machine learning, but they can't scale them for the business, let alone across the enterprise, which is where everyone needs to get to. And that's where we're really focused on the offering set. So from a storage, that would be where we do the AI. And of course, the third one is just containers in general, which by the way, they intersect because a lot of things you're going to do in containers you're going to intersect with AI and especially when you go multi-cloud, but there's a whole different, hey, let me modernize my application infrastructure and it's a different conversation with clients. So we see people being very successful and that's where you're seeing from a storage development investments is we're going into that direction, helping clients with those different environments. Why is AI at scale so difficult? Is it the silo, data silo problem? So first of all, there's a little miss about AI. The one is a black box, it's mysterious, it's really just computer science. I mean, it's a process, especially when you're talking about machine learning, deep learning, it is you're doing stats, you're just driving it, you're being iterative with it, using customized silicon to do it faster, like GPUs. The other one is it's easy and it's not. So in infrastructure matters, so when you get going, what we see is people just give a particular data scientist an environment, just to say you bring the data in you need and we'll help you with the governance strategy, but it's typically getting something to just be valuable to the business. Then what happens is they see another random act of AI or another area that, and it becomes data silo, but they're trying their best to stay away from the data scientist, given the right support, which I think is the right thing, because the biggest thing is not have it all controlled and centralized, you want to let the business units drive. But then what happens is you have almost like data warehousing in the past, right? You have these islands and now you don't have any trusted true source of the truth and you don't have ability to get, you force everyone to do the hard bit. The hard bit is actually having the right data, do all the 80% of cleansing that data, having the right governance and security about that, and then you're adding to over time. What if you do a couple applications, and it's not the first one you do, but once you do the second, third, maybe fifth, definitely by the 18th application, you want to bring that together in a shared platform and that's where IBM Storage plays, and that's where we're truly differentiated compared to the storage industry. So we have assets that no one else has, like what we do is spectrum scale, where we can literally scale up from just an individual server to half rack, and we can take the same environment and do the largest AI sub-computers in the world. But the key thing is you need, you can't have AI without IA, information architecture, and that's on the software side, but it definitely has to be in the infrastructure. And we're doing it across hybrid multiclouds, so we're doing that on-prem, but trust me, AI is absolutely in the cloud, so we can extend those environments and run the same thing in any of the public clouds. And what's the storage enabler? Is it software defined, is it, you mentioned architecture, what is the linchpin there? Well, so one, it is a software defined, so in this particular area where we're helping people is our file system called Spectrum Scale, but it allows us to do from the very small to the largest environments, right? That allows you to scale, and it also runs all different assets, so it's unstructured. So you're able to run Hadoop native Spark, but you're able to file, you're able to block, you're able to bring it together, and you're able to start small, but you're able to scale and keep up. The thing about AI is you go from, I want to collect the information, get after it. You also need metadata, so we have products like Spectrum Discover to show you the metadata so we can actually track it. So it doesn't become the junk drawer in the sky that we've seen with a lot of data lakes. But then it's interesting, you have to go in to actually do the training, and that's where you're using custom silicon GPUs. That dramatically changes the performance you require. So these GPUs used to run at 60 gigabytes a second. Now they're 150 gigabytes a second. We know storage, traditional storage that you get from the environment, we might get from PureNet app or EMC, they run at sub 20 gigabytes a second. So how do you do this? It's a different architecture. It's actually based upon a true scale out what we've seen in the largest supercomputers in the world, but you're able to bring that environment so we can actually do, bring in all the data works, get under one governance and strategy, but then you can actually keep up with the performance of the true inferencing and driving GPUs. Now once you have a trusted source, you can scale this out literally the largest AI supercomputer in the world. So we can show you scale on the exact same components, start off a half rack and go to the biggest thing that's in the world. But the key thing is you need to actually have the performance. So if you have this data backplane, if you call that, now people will still take a lot of the data, they'll bring it into the servers and they'll crunch it with GPUs. So they say, well, okay, your storage doesn't need to have that performance. But what you'll find is once you have a common backplane, which is how IBM did it, now you have different business units, almost hub and spoke grabbing the data, but they have one true source of data. They're able to get after it, they're able to get their other data that other groups are looking for, but now they're able to now scale it into the enterprise. Because some of it is just an AI call, right? API call. So you're doing applications and they just want to do an API call to the same environment. And those have to be fast because now you're inferencing, so it's sub seconds, but you need that performance. So what you need is by bringing it all together, you can either do data silos, which are easy upfront, but by the time you do the second and third, you're doing all that same work over and over and you don't have a trusted source, you got to bring it together, but you can't bring it together in any storage. And we don't bring it together on the same storage we do for VMware environments or others, it is different storage. And it's made specifically for this environment. And it's something that actually IBM is leaps and bounds above everyone. In fact, at Supercomputer, they do these type of analysis and we're like two X are best competitor in benchmarks. By the way, that competitor uses spectrum scale. But it is a different architecture. But if you don't put it together, and I would also say that when you get started, no one starts with the big. In fact, that's almost a mistake. What you want to do is let the data scientists have the creativity, drive it, get business outcome. But then you need to be thinking ahead, how do you bring it together so you have a shared? Because again, the way you're really going to drive it across the enterprise, all the different processes is actually having some API calls coming in, which are not going to be their own environments altogether. You make sense? Yeah, you've mentioned a couple of times infrastructure matters and I want to tie that into the 80%. It was at this very venue in 2009 when Paul Moritz said, as the CEO of VMware, we're going to run any workload, any application virtualized. And a lot of people were skeptical. And I remember the time thing, well, mainframes kind of did that. And now you're talking about the 80%. And VMware, I think, largely proved that you could run that at high performance, at least adequate performance. Now, you're seeing a similar discussion around cloud, but it's somewhat different because of some of the things that you were just mentioning. What does that world look like? Obviously, hybrid fits in. You mentioned the red hat acquisition. That's key part of IBM's go forward, multi-cloud strategy. So should we think about it? What's similar and what's different than the sort of VMware virtualization, mainframe virtualization days? Okay, that's interesting. And then you're tying into the cloud adoption as well. And we do think in this IDC, about 20% of the workloads have gone, about 80% are waiting. And I never thought of that way. It's a good analogy to virtualization. The easy stuff went first. And then what you have to do is have the databases. Remember, that was a big issue. You couldn't do that for years. But then all of a sudden you move it. It's like, why wouldn't you do it? But what's happening is what we're seeing is this the mission-critical workloads. And they're either regulated industries, but it's for different reasons. They're running it in different places. But it might be a security concern. It might be a scalability. It might be regulated industries. But there's reasons, or maybe you have to refactor the applications to actually run cloud-native. Because the lift and shift wasn't the same economics as you thought. So what we think is the next 80% is not going to be led by the application, or things like Office 365. We actually think it's going to be people building their real mission-critical workloads. And that's a different conversation. That's where we think the market's kind of playing to what we do at IBM and infrastructure, where you need to have the technology, but also the expertise and industry to move it. And then security becomes key concern. Yeah, so we mentioned Red Hat here. And I know you can't say too much, but we know about kind of the cloud-native modern, multi-cloud stuff. But Red Hat also has quite a bit of a storage portfolio. Seth and Gluster acquisitions open source. Wondering what you can just, as an IBMer, tell us about what you think of that portfolio. So we can't talk about what's going to happen afterwards, but also I think we made it very clear. We believe Red Hat is, and you used VMware, I think it's a good analogy. We're going to keep it, we're going to be independent, they serve a role, we're not going to change it. And I think that's a very important part of the message. But we can't talk about the assets. And I did my own kickoff to my team, my partners. And I used, I kid around, that was a storage acquisition. So it was my $34 billion, right? So, but I think it has a good play. It's completely complementary to what we do. They have some great technologies, but also we bring things to them that are interesting. But a conversation, when someone's trying to say, where am I going? And for strategy, we believe containers are critical. We believe Linux is critical. If you look at what people are doing on the cloud, it is, it's already crossed, it's mostly Linux, and the enterprises have chosen Red Hat. Now, if you think of what we can do with that particular environment, to have the conversation make relevance about what we can do to help you on-prem, but now you can run the same thing on-prem, you can put it literally anywhere. Now that's a strength of what Red Hat would bring us. On a storage side, they have great assets. I'm kind of salivating to help them out with that. Now, what they don't do is some of the things that we can add to them, right? So, and I don't think we're commenting on any roadmap. I think what they have and what we have is, to be honest, complimentary, if that makes sense. Well, and I think, you're familiar with our old true private cloud, no, it's evolving to true hybrid cloud. And what you just described is true hybrid cloud. Run it wherever you want. You're agnostic to where it runs, but you want that cloud operating model to be the underpinning of the experience. And you don't want to be locked in. So that's where I think if you look at hybrid, you want to make sure, I think Red Hat allows us to do that. I think IBM is synergistic to it. I think we can bring a lot of, how do you bring the integration capabilities that IBM brings on applications and help these mission-critical environments that need some industry expertise to do that, to bring them to the cloud itself. And maybe it doesn't matter where on the cloud. And one of the things we were commenting on the open is the hybrid, multi-cloud world. It's complicated, and IBM has a strong history with services to help drive that. Give us a little bit of your insight as to what IBM brings to kind of that multi-cloud environment. It's almost too much, right? So what we're doing is really working on the overall. How do we simplify it? So we're going to meet the client where they are. So everyone's at different, you have to almost find out where they are in the journey. And then even a particular client, you'll find different business units on different parts of the journey. So we can help them anywhere from helping figure out architect where they would go. We can help them move to the cloud. We can actually help them manage the cloud. And they're also going to meet them where they are. So we have, if you think about what we can do with AI, we have full AI stacks and enterprise capabilities, but some people choose to just use their own open source. And we can help them. In fact, you see our multi-cloud manager allows you to manage regardless of what you've done for your build environment. So what we're going to do is meet clients where they are and help them do the last mile. And then with service and support, we're able to, if you look at what we do with GTS, we're the largest Red Hat support organization. If you look at what we do with GBS, we can help people build out their own platforms and give an overall restriction and how to go drive AI at scale in their environments. So I think it does play to us. And I think the Red Hat acquisition just adds, I think one-on-one is three. So. Well, and I think, you know, we were talking in our open that just even IBM's giant application modernization opportunity, right? I mean, because we tend to think about, you know, AI and leading edge, but there's just so much modernization opportunity. And a lot of guys, you know, they don't want to go it alone with open source. The fact that IBM's there, you know, with that big blue blanket, I called it, say, okay, we're going to help you through your modernization, you know, initiatives. You know, we think a big deal. You know, we're excited about that next chapter. And think about Red Hat. Red Hat does a lot of consulting, but they have been disciplined. They help you get rail going. Once you start docking open stack, right? So, I was going to be, open shift, excuse me. Now that is a whole different way to look at your environment and your applications. And that takes a higher level. So it's like one-on-one is three. They don't do a lot of that now, right? Well, and it's instant developers, too. That's the other thing we said just that. Well, how many developers have Red Hat at 8 million? I mean, we were talking about Cisco before, with, you know, half a million, which is great, but you're talking about 8 million. And, you know, despite IBM's efforts around, you know, blue mix, so that was a heavy slog. Now, all of a sudden, you've got 8 million developers. It is, look, Red Hat Linux is running the internet. You know, we know this. So that's exciting. I want, my last question, Ed, is you've made a career, early part of your career, and taken startups, and getting them to a point where they could be acquired, and a very, very successful career there. Then you joined IBM, spent some time at MIT, you know, getting even smarter. When you sit back and look at the industry, the storage industry in particular, somehow, despite the trend toward cloud and bigger is better, you still see these specialists, you know, popping up all over the place. You see guys like Pure, you see guys like Nutanix. And, you know, we saw that earlier with the compelents and the three-pars and the isolans. We said, okay, maybe this is the last wave. But somehow, storage innovation continues to occur. Do you think we're seeing sort of the end of that, sort of storage startup craze? Can independent storage companies continue to survive? What are your thoughts as an industry observer? So I think it's more difficult, but there's plenty of innovation. So you're seeing it as we just help people get to their journey. You're going to see different technologies at even IBM. You're going to see our portfolio change rather dramatically to keep up with the trends and you need to do it. So I don't think it's over with. So I never want to be quoted that I think there's no innovation left and there's a role for it. You saw storage, you know, grow last year, right? So it was, there's always been growth areas, but it's been flat. Also, it really took off. And a lot of that is because what we're doing on AI, which is not your average, it's definitely not what Pure does as far as for storage, right? But you're going to see, I think you're going to see innovation, I think you're going to see that continue. But I think it's harder and harder for these independent firms, mostly when they scale. I think it's the innovation piece and I think you're seeing guys like us and EMC and others innovate very quickly and if our innovation is speeding up with investment because we have to, our clients are demanding it. And the VCs keep pouring money in. I mean, you're seeing that in the data protection space and the others. Data protection is now cool, right? So we see them all the time. I think in general, if you look at data protection, it becomes now your secret weapon when you're talking about being data-driven in a classical environment. You can get copies of your data via APIs anywhere in the environment. So I think it's a really big play. So it opens up new opportunities beyond just backup, right? For whether it's DevOps or maybe disaster recovery or ransomware, even analytics because the backup corpus has all the data in it. So a lot of possibilities. Ed, thanks so much for coming to theCUBE. We're going to see you also on Wednesday. And looking forward to that. So thank you. Thank you. All right, keep it right there, everybody. We'll be back with our next guest. You're watching theCUBE from day one at IBM Think in Moscone. We'll be right back.