>> Hi, welcome once again to Wikibon's weekly research meeting from The Cube's Palo Alto studios. (upbeat electronic music) >> I'm Peter Burris, and we're being joined as always by Wikibon's team of analysts, including George Gilbert here in the studio with me. On the phone we have David Floyer, Neil Raden, and James Kebyouis. And today, what we're going to do is we're going to talk about some of the lessons that we learned in 2008 or 2017. Over the course of the next month, Wikibon is going to put a fair amount of research into making our annual predictions, and this is the first step. What lessons did we learn? What is working? What isn't working? As a consequence of some of the things that were tried, predictions that were made, and initiatives that haven't necessarily panned out. Now the reason we want to do this is not just to talk about technology, but we're trying to bring the idea to those users out there who are in the midst of budgeting, about where they should continue to place bets, and where they might want to start thinking about rationing down; things that don't seem to be panning out. So there's a lot of ground to cover, and let's get started. And I want to start with you David Floyer. So the first thing I think we've learned in 2017 is that the cloud is not going to be homogenous. Do you agree? >> David: Absolutely, it's becoming very, very heterogeneous. We brought into play, the concept of true private cloud, and we're seeing that develop very strongly, and we're predicting that again. In the future we'll develop service at a completely different way of doing storage that's coming from the cloud. From the hypervisor cloud, into the private cloud, and for general purpose. And wishing really some very big changes in how systems will begin to be developed. >> So with that as a basis for some of the kind of Macro Trends, the idea that business is not going to move to Cloud as we like to say. The cloud is going to move to business. There are a number of applications that are driving some of these changes. Neil I want to start with you. One of them is big data, or perhaps we should finally start calling it analytics. What is it about analytics that is starting to catalyze a rethinking of the overall architecture that we're going to use to sustain some of these digital business changes, that all companies, all institutions face? >> I don't know how it started Peter. People have been doing analytics for decades while Corporate IT was more or less obsessed with operations. But over the last five to ten years, analytics has just become the most important thing, and it's not a flip flop. The problem is the approach to analytics has jumped from one thing to another so quickly. I don't think that anyone has had a chance to really perfect their approach. We went from predictive analytics, and then we went to data science and big data. And now everything is machine learning and artificial intelligence. If I were inside an organization right now, my head would be spinning. So we'd have to elucidate some clear directions for people about what works and what doesn't, and what the level of effort in as it spended to get things done. >> So (mumbles)? Is it safe to say. Is it face to say. Is it safe to say at this point, that the kind of general purpose notion of big data, where you throw everything into a single store, like a datalake. And then you have everybody run around looking for data. Is that starting to break down and become increasingly specialized? Is that kind of what we learned in 2017? >> I think it's safe to say that big data never really crossed the Cazim. Its closest application to something that would be appealing to mainstream customers, was taking ATL and offloading it in front of very expensive dated warehouses. But the way the open source ecosystem, principally with the (mumbles) distributions that curated all these open source components. The way it tried to attack that problem was so complicated in terms of the administrative demands, that most customers choked on it. >> Peter: So we're seeing increasing specialization, in part because of the nature of the problems that people are trying to solve, but also the complexity of the underlying solution. So that leads to a third question, and the third question is, we talked about cloud not being homogenous. We talked about big data becoming more specialized and solution oriented, outcome oriented. One of the other big drivers in all this, David Floyer is IOT. We'll talk in a second about how IOT and analytics are going to come together. But what are we learning from IOT in 2017. >> What we're learning, is that the edge is again, not homogenous. And it's much better to look at the, break out the edge, and break out the IOT at the edge, into a primary layer, and the secondary layer. The primary layer is the layer that is a solution which takes the sensors, takes equipment, takes AI technologies, and brings them all together, add a solution to a business problem. And we believe that that is a much lower cost, and volume approach to the problem, then everybody, every IT making their own equipment in their own factories or entrances So the primary is where most of the data is going to be generated, and also where most of the generated data is going to be compressed down from, maybe as much a million to one into the secondary layer. And that's the interface between the primary layer and the cloud computing, whether it be (mumbles) by the cloud, or public cloud, or any combinations of those. That's the tertiary layer, and the secondary level would be that interface at the edge between the primary devices, and the cloud computing the rest of the enterprises, dependent upon. >> So Jim Kobielus, we've got three lessons learned in the table. Clouds are homogenous, analytics are increasingly going to be a feature of applications. But that's going to require to be retooling. IOT is not going to be homogenous, it's going to drive new data sources, and new opportunities to create value in (mumbles). Where are developers in all this. What do we learn, or what are we learning as the developer committee starts to try to participate more in the process of creating new levels of digitally based value in business. >> Right, what we're going to need to. Well what developers are learning, and enterprises are learning, is that their current group of core developers are not prepared for this AI at the edge revolution. Not prepared in terms of skill, the tools at their disposal. The DevOps pipeline, their workflows that are in place. The teaming arrangement in collab are still dataalkes themselves. Not prepared to do AI effectively, or drive it effectively to the edge, where it can achieve the intent that (mumbles) who (mumbles) in terms of business value. So what that means is, in 2018 and beyond, if you're an enterprise IT manager, you're an analytics manager. Where do you place your budget? Is it skills up where you hire the right people? Do you operate your tools, and somehow make due with the DevOps tools you have? I would bring more of the, for example model governance over algorithms and deep learnings and (mumbles) model into the core governance structure you have. Can you do a datalake? Do you have datalakes that are architected to handle machine data in great volume Like (mumbles) and exabytes of machine data generated by all these end points. Okay, there's all these decisions that need to be made, and there's money that needs to be spent to invest in this entire development infrastructure ecosystem, to really prepare yourself to build these disruptive applications that might take your industry by its storm. None of this comes cheap. >> Peter: So it seems guys like, we're in a situation where the technology in many respects is available to undertake and build, and deploy, and generate value out of some of these new classes of applications. But skills are very, very unevenly distributed. Neil Raden, let's talk a little bit about that. What is the core skills challenge that businesses face today as they attempt to explore new ways of solving problems with digitally related technologies. >> I think that software vendors are going to provide a tiered capability, just like we've seen in other kinds of analytical tools. Where you have a small number of people at the top of the tier, who have the background and the skill to understand whether this model was the appropriate model, or whether we found a correlation that was serious, because they were all time series, or something like that. And then you have a larger group of people who use these tools to drive a machine learning algorithm, or like data robot, where it just runs 10 or 12 different algorithms, and it helps you find the best one and so forth. But that doesn't mean that it's correct, and that doesn't mean that those people understand the statistics that are generated by the model. That requires governance of the people at the top of that tier. And then of course, there's the lower tier, which is how they communicate to these people, what you've done with these techniques. >> Peter: So this is a broad problem, it sounds like. It sounds like we've got a skills deficit problem that's going to have far reaching impacts. We'll talk more about this during predictions, but I think there's one that's on everybody's mind right now. Are we going to se specialist software and solution vendors emerge out of this to start the process of at least solving some of these problems, and showing the industry how to go about it. Or is this something that all large enterprises, and mid-size enterprises are going to have to do on their own, and they got to start throwing an enormous amount of money at these issues? David Floyer, give our CIOs a kind of a vision of where they should be thinking right now about how to address the challenges of skills. >> David: Well, the big decision to make for all enterprise, but most enterprises, is whether to, the degree to which they should invest in their own solutions. Their own AI solution. Or should they wait until those solutions are included in (mumbles) the packages, and general purpose packages, and packages they get from (mumbles), then the (mumbles). And if you're a very large enterprise, and you can see a clear business differentiation, then clearly that investment can be justified. But I think for many enterprise CIOs, they will sit back and wait, and see the degree to which they need to invest. I don't mean to say that they should be actively seeing what is available in a marketplace, but they should be probably spending more time reaching out to potential vendors with a solution, who can generate volume, rather than trying to create snowflakes on their own. >> Peter: So, before we get to the action item round, Jim I want to build on that very quickly. So Dave's arguing essentially that we're moving into a buy vs. build as we go through this transformation. I think we all agree, that's where we are today. Next question though. Is it going to be buying software, or is it going to be buying services? Or some combination of the two? What did we learn in 2017, about how the availability of increasingly advanced services, especially in the AR realm from some of the big cloud suppliers, is changing or altering the way businesses think about how they're going to generate value out of these technologies. >> Jim: Yeah, I think right now what we're seeing is the swing is towards buying services. Buying cloud services that have machine learning, deep learning, AI, (mumbles) again from the usual suspects AWS, Microsoft. (mumbles) has been Google and IBM, so forth. What we see right now in the whole developer war, is to win the hearts and minds of AI developer, is it's coming down to whose cloud are you going to put your data in. You can do your model training & development and deployment. Whose framework? AWS's mxnet? Microsoft's CNTK whatever? Google's Tensorflow? Are you going to use in those bendors, the solution providers in those frameworks provide free training models, and a lot of other capability to not build out, not only the models, but to provide a floor DevOp pipeline for the data (mumbles) that you have to be standardized on one solution provider a lot more more than others. >> Peter: George, George. Hey Jim, let me bring George in. George, what do you have to say about (crosstalk). >> I think we've seen this. We've seen this movie before when enterprises started to build out their applications. At one point they were thinking of large enterprises, custom data modeling how their entire enterprise worked, and realized they didn't have the skills to do that. They brought (mumbles). So I don't think the choice is binary between buying services, or buying apps. I think there's also, are we going to wait for the install base of apps. The big vendors who've installed the large horizontal apps to add machine learning capabilities to those applications, where we start to surround those legacy apps with more niche package solutions. And then the third one is, will we see vendors like IBM, and maybe Accenture, which have a mix of services and some repeatable IP. >> Great, so the one I'll add to this before we do the Action Item guys, is I think one of the more important things that we're facing in the industry right now. Is, as it becomes evident per David's earlier point that the cloud is not going to be homogenous. Are we moving into another round of platform wars? Where users have to be very, very smart about what platform they choose. Yes, but increasingly having the options to do the appropriate level of integration across whatever arrangement of cloud services, on premise, true pride of cloud, etc. Probably something. A lesson that we've learned, and one that our clients will increasingly tell us that we have to focus on. Okay, Action Item round guys. David Floyer, I want to talk with you. David Floyer, action item. >> The action item for me is actually an infrastructure. There is a tremendous opportunity evolving to develop, be able to put applications with far more data on to their systems. And those are based on a change in architecture, which we're calling (mumbles), which is tripping away the storage and the networking completely from the processes. Being able to assemble systems which do things, which are just unimaginable, just by the (crosstalk). >> George Gilbert, action item. >> I'd go back to picking how you're going to divide your efforts among extending your existing package ups with machine learning capabilities, and finding where the highest auto y areas for those are. Look at the emerging, sort of, I don't want to say startups, but younger companies that are adding these complimentary capabilities, and. >> Peter: Okay, good. Next, Jim Kobielus, action item. >> Yeah, well action item is explore the new generation of high level developmental traction framework for AI and deeplink light. The new glue on framework that Microsoft and (mumbles) released a couple of weeks ago. That will enable the rest of us developers to be able to do deep learning AI development using code and visual paradigms that they have grown to love and use in their core development initiative. >> Peter: Neil Raden, action item. >> I like machine learning even though it has a lofty title that maybe it doesn't deserve. It's not that complicated. But more importantly, it creates opportunities for organizations to do things that really can help them. I think we spend too much time talking about AI, and I think the average organization needs a computer that thinks like a human being, about as much as we need airplanes that flap their wings. But there's too much time on AI, which is a very esoteric area. Facial recognition and all that other stuff. That's going to be packaged with things if you need it, but companies don't need to worry about finding people who can develop that. >> No need to anthropomorphize what doesn't need to be anthropomorphized. Okay, so here's our overall action item. 2017 has been a year of significant success in the computing industry, as businesses increasingly woke up to the idea that the transformation to digital business is not just about taking costs out of IT. It's about doing things differently, and specifically doing more with data. We've seen a lot of leaders in this realm. Companies that have been called digital natives have paved the way, but a lot of other industries are now recognizing that the role of data as an asset, is crucial to their future. And they want to find ways of appropriating that. In particular, we think that there are three lessons that have been learned at the technology level. Lesson number one, the cloud is not going to be homogenous. The cloud is going to be a combination of technologies, each optimized to handle data, as it pertains to particular uses, application forms and workloads, in the natural and appropriate way. Data will drive workload, will drive, cloud implementation. Number two, is that one of the key issue, or one of the key areas of that changes the transformation from big data concepts to analytic practicalities. We've got years of working with analytics. The technology is improving, the hardware is improving, and now we can apply new and interesting ways. And very importantly, that includes applying it to existing legacy applications to extend their useful life as well. A lot is going to go on into this, but the good news, ultimately, is that technology is becoming increasingly usable and increasingly useful to business. Third, the IOT, or internet of things, is going to have an enormous consequence in how we consider the arrangement of IT assets, IT investments, and IT personnel. And our expectation ultimately, is that that will continue to be a crucial determinant of the decisions that ultimately get made, if success is a criteria. Because our observation is, yes software is going to eat the world, but it's going to eat it at the edge. The last poin that we want to make here ultimately, is that a lot of IT organizations have to fess up to the reality that they're not skilled to do a lot of these things. They're not skilled to fully support the business's needs in these transformations. We are no longer in control of the speed of transformation in our industries, that's being set by our competitors who may be better or worse than us at introducing some of these new technologies, and taking advantage of them. And introducing a new business model and customer experience capabilities. As a consequence, there's going to be a new round of value being created by solution providers, utilizing different cloud options, different IOT options, and different AI options in response to expertise about how those solutions need to be deployed. And IT has to accept that, sooner rather than later, and start the process of establishing the frameworks for strategic management of those suppliers, so they can appropriately weave them into