 Live from San Francisco, it's theCUBE. Covering IBM Think 2019, brought to you by IBM. Welcome back, we're here at Moscone, north, you're watching theCUBE, the leader in live tech coverage. This is day four of our wall-to-wall coverage of IBM. Think the second annual IBM Think, first year at Moscone. Dave Vellante here with Stu Miniman. Eric Herzog is here, he's the CMO of IBM Storage and Sam Warner is the VP of offering management for storage software at IBM. Guys, welcome back to theCUBE, always good to see you both. Thanks. Thank you. So we were joking yesterday and today, of course, multi-cloud, the cloud's opened, it's been raining, it's been sunny today, so multi-cloud is all the rage. Evidently, you guys have done some work in multi-cloud, some research that you can share with us with? Yeah, so a couple things, first of all, the storage vision in multi-cloud at IBM for years. We work with all the cloud providers, including IBM Cloud, but we work with Amazon, we work with Azure, we work with Google Cloud, and in fact, our Spectrum Protect, modern data protection product, has about 350 small and medium cloud providers across the world that use it for the engine, for their backup as a service. So we've been doing that for a long time, but I think what you're getting is, what we found in a survey multi-cloud, and I actually had a panel yesterday and all three of my panelists, including Aetna, use a minimum of five different public cloud providers. So what we're seeing is, hybrid is a subset of that, right, on and off, but even if someone is saying, I'm using cloud providers, they're using between five and 10, not counting software as a service, because many of the people in the survey didn't realize software as a service is, theoretically, a type of cloud deployment, right? So that's obviously not just the big three, or the big five, we're talking about a lot of smaller clouds, some of the guys, maybe that are using your Spectrum Protect for backup, local cloud providers, right? And then add SaaS to that, you could probably double or triple it, right? Right, well, we've had been very successful with SaaS providers, so for example, one of the people on the panel, a company called Follett, they're privately held, in the mid, close to a billion dollars, they provide services to universities and school districts and they have a software package for universities, for the bookstores, and to manage the textbooks, and another software is a service for school districts across the United States. They have 1,500, and it's all software service, no on-prem licensing, and that's an example, that's, in my mind, that's a cloud deployment, right? Ginny talked Tuesday about chapter two, how chapter one was kind of the, I call it commodity cloud, but apps that are customer-facing, chapter two, a lot of chapter two anyway is going to be about hybrid and multi-cloud. I feel like to date it's largely been not necessarily a purposeful strategy to go multi-cloud, it's just a multi-vendor. Do you see customers actually starting to think about a multi-cloud strategy, if so, what's behind that, and then more specifically, what are you guys doing from a software standpoint to support that? Yeah, so in the storage space where we are, we find customers are now trying to come up with a data management strategy in a multi-cloud model, especially as they want to bring all their data together to come up with insights. So as they start wanting to build an AI strategy and extend what they're doing with analytics and try to figure out how to get value out of the data, they're building a model that is able to consolidate the data, allow them to ingest it, and then actually build out AI models that can gain insights from it. So for our software portfolio, we're working with the different types of service providers, we're working closely with all the big cloud providers getting our software out there and giving our customers flexible ways to move and manage their data between the clouds and also have clear visibility in all the data so they can bring it together. You know, I wonder sort of what the catalyst is there. I wrote an article today, it's going up on SiliconANGLE later, and I talked about how the first phase was kind of tire kicking of cloud, and then when the downturn hit, people went from CapEx to OpEx, it was sort of a CFO mandate, and then coming out of the downturn, the lines of business were like, whoa, agility, I love this, so shadow IT, and then IT sort of bought in and said, we got to clean up this mess. And that seems to be why at least one catalyst for a company saying, hey, we want a single data management strategy. Are you seeing that, or is there more to it? Well I think, first of all, we're absolutely seeing it, and there's a lot of drivers behind it. There's absolutely IT realizing they need to get control over this again. Governance, compliance, security, edicts. And think about all the new regulations. GDPR has had a huge impact. All of a sudden, these IT organizations need to really track the data and be able to take action on it. And then you have all these new roles and organizations, like data scientists who want to get their hands on data. How do you make sure that you have governance models around that data to ensure you're not handing them things like PI? So they realize very quickly that they need to have much better control. The other thing you've seen is the rise of the vulnerabilities. You see much more public attacks on data. You've seen C-level executives lose their jobs over this. So there's a lot more stress about how are we keeping all this data safe. You're right, boards are getting flipped, and it's a big risk these days. Well the other thing you're seeing is legal issues. Canada, the data has to stay in Canada. So if you're a multinational and you're a Japanese company, all your Canadian offices, the data has to be some cloud provider that's got an office in Canada. So if you're a Japanese headquarter company and you're using NTT cloud, then you've got to use IBM or Amazon or Azure because you have to have a data center inside the country just to have the cloud data. You also have pure maturity in the market. I would argue the cloud used to be called the web and before it was a web it was called the internet. And so now that you're doing that, what happens is in the bigger companies, Procurement is involved, just the way they've been involved in storage servers and networking for a long time. Great, you're using Cisco for the network. You did get a quote from HP or using IBM storage, but make sure you get at least one other quote. So as that influences aside from definitely getting the control is when Procurement gets involved, everything goes out for RFP or RFQ or a tender as they say in Europe. And you have to have multiple vendors and sometimes you may end up for purely, we need the way to club them on price. So we need IBM cloud and Microsoft so we can keep them on us. So when everyone rushed to the cloud they didn't necessarily do that, but now that it's maturing. Yeah, it's a sign of maturity. It's a sign of maturity that people want to control pricing. All right, so one of the other big themes we've been talking a lot about this week is AI. So, you know, Eric talks about when we roll back the clock, you know, I think back to the storage world, we've been talking about intelligence and storage for longer than my career. So Sam, maybe you can tell us what's different about AI in storage than the intelligence we've been talking and, you know, what's the latest about how AI fits through the portfolio? Yeah, that's a great question. And actually a lot of times we talk about AI and how storage is really important to make the data available for AI, but we're also embedding AI in our storage products. If you think about it, if you have a problem with your storage product, you don't just take down one application, you can take down an entire company. So you've got to make sure your storage is really resilient. So we're building AI in that can actually predict failures before they happen so that our storage never takes any outages or has any downtime. We can also predict by looking at behavior out on the network, we can predict or identify issues that a host might be causing on the network and proactively tell a customer before they get the call that the applications are slowing down and we can point out exactly which host is causing the problem. So we're actually proactively finding problems out on the storage network before they become an issue. Yeah, and Eric, what is it about the storage portfolio that IBM has that makes it a good solution for customers that are deploying AI as an application in use cases? Yeah, so we look at it up. So what is AI in the box, if you will, in the array and we've done a ton of work there, but the other is as the underlying foundation for AI workloads and applications. So a couple of things. Clearly AI often is performance dependent and we're focused on off-lash. Second thing is Sam already put it out, resilience and availability. If you're going to use AI in an automotive factory to control the supply chain and to control the actual factory floor, you can't have it go down because they could be out tens of millions, hundreds of millions of years for just that day of building Mercedes or Toyotas or whatever they're building if you have an automated factory. The other areas, we've created what we call the data pipeline and involves four members of our stored software family, our spectrum scale, a highly parallel file system that allows incredible performance for AI, our spectrum discover, which allows you to use metadata, which is information about the data to more accurately plan and the AI software from any vendor can use an API and go in and see this metadata information to make the AI software more efficient that they would use. Our IBM Cloud Object Storage and our spectrum archive, you have to archive the data, but easily bring it back because AI is like a human. We are smart humans are learning nonstop, whether you're five, whether you're 25 or whether you're 75, you're always learning. You read the newspaper, you see of course the cube and you learn new things, but you're always comparing that to what you used to know. Are the Russians our friends or are Emmy? It depends on your point in time. Do we love what's going on in Germany? Depends on your point in time. In 1944 I'd say probably not. Today you'd say what a great democratic country, but you have to learn and so this data pipeline, this loop, our software is on our storage arrays and allows it to be used. We'll even sell the software without our storage arrays for use on any AI server platform. So that software is really the huge differentiator for us. So can you, as a follow-up to that, can you address the programmability of your portfolio, whether it's through software or maybe the infrastructure as well. I'm thinking infrastructure as code. You mentioned APIs. You mentioned the ability to go into like spectrum discover, for example, access metadata. How programmable is your infrastructure and how are you enabling that? Yeah, I mean, across our entire portfolio we build restful APIs to make our infrastructure completely extensible. We find that more and more enterprises are looking to automate the deployment of the infrastructure and so we provide APIs for programming and deploying that. We're also moving towards containerizing most of our storage products so that as enterprises move towards Kubernetes type clusters, we work with both Red Hat and with our own ICP and as customers move towards those deployment models and automate the deployment of the clusters, we're making all of our storage available to be deployed within those environments. So do you see an evolution of the role of a storage admin from one that's sort of provisioning lungs to one that's actually becoming a coder, maybe learning Python, learning how to interact through APIs, maybe even at some point developing applications for automation, is that happening? I think there's absolutely a shift in the skills. I think you've got skills going in two directions. One, in the way of somebody has to administer hardware and replace parts as they fail. So you have lower skilled jobs in that side and then I believe that yes, the people who are managing the infrastructure have to move up and move towards coding and automating the infrastructure. As the amount of data grows, it becomes too difficult to manage it in the old manual ways of doing it. You need automation and intelligence in the storage infrastructure that can identify problems and readjust. For example, in our storage infrastructure, we have automated data placement that puts it on the correct tier. That used to be something a storage administrator had to do manually and figure out how to place data. Now the storage can do it themselves. So now they need to move up into the automation stack. So we've been talking about automation and storage also for a lot of years. Eric, how are enterprises getting over that fear that either I'm going to lose my job or this is my business we're talking about here. How do I let go and trust? I love I saw downstairs, there was in the automation booth for IBM, it was free the humans. So we understand that we need to go there. We can't not put automation with the scale and how things are moving, but what's the reality out in the field? So I think that the big difference is, and this is going to sound funny, but the economic downturn of seven, eight and nine. But that downturn hit and was certainly all over the IT press. Lay off, lay off, lay off, lay off, lay off. So we also know that storage is growing exponentially. So for example, if I'm Fortune 500 company X and I had 100 people doing storage across the planet, if I laid off 50 of them and now I'm recovered I'm making tons of money, my IT budget is back up. I didn't go to the CIO and say, you can hire the 50 storage people back. You can hire 50 people back, but no more than five or six can be storage people. Everything else has to be DevOps or something else. So what that means is they are managing ungodly amounts of more storage every year with essentially the same people they had in 2008 or maybe a tiny bit more. So what matters is you don't manage a petabyte or in the old days, half a petabyte. Now one storage admin or backup admin or anyone in that space, they want you to manage 20 petabytes and if you don't have automation, that'll never happen. Stu and I were interviewing Stephen Hill from KPMG yesterday and he was talking about the macro numbers show we're not as globally and even in the US, we're not seeing productivity gains. I'm saying, yeah, you're not looking at the storage business because if you look at anybody who's running storage they're doing way more with much less to your point. So for example, when Sam talked about our easy tier, we can tier not only is the AI based. So in the old days, when you guys weren't even boring yet, when I was doing it, what was it? It was move the data after 90. So first was manual movement. Then it was set up something, a policy, remember policy automation was the big deal 10 years ago. Automatically move the data when it's 90, 60 or 30 days old. AI based, what we have an easy tier, automatically will determine what tier it should go on whether when the data is hot or when the data is cold. And on top of that, because we can tier over 440 arrays that are not IBM logo, multi-vendor tiering, we can tier from our box to an EMC box. So if we have a flash array and you've got an old or all hard drive that you've moved into your backup and archive tier, we can automatically tier to that. We can tier from the EMC array out to the cloud. So, but it's all done automatically. The admin doesn't do anything. It just says source and target and then the AI does all the work. That's how you get the productivity that you're talking about that you need in storage and backups even worse because you got to keep everything out. Stan mentioned GDPR, all these new regulations in the federal government, it's like keep the data forever. So in that case, the machine can determine whether or not it's okay to put it in the cloud if it's in Canada or Germany or wherever. The machine can adjudicate and make those decisions. And that's what the AI, so in that case, you're using AI inside of the storage system versus what we talked about with our other software that makes our storage systems a great platform for other AI workloads that are not, if you will, AI for storage, AI for everything else. Cars or hospitals or resume analysis, that's what the platform can, but we put all this AI inside of the system because there aren't that big, giant global Fortune 500 has 55 storage admins and in 2007 or eight, they had 100, but they quintupled the amount of storage easily if not 10X'ed it, so who's going to manage that? Automation. Guys, good discussion, not everyday boring old storage, talking about intelligence, real intelligence this time. Eric, Sam, thanks very much for coming to theCUBE. Great to see you guys again. Thank you. Hey, you're welcome. All right, keep it right there, everybody. Stu and I will be back with our next guest shortly, right after this break. John Furrier is also here, IBM Think. Day four, you're watching theCUBE, right back.