 Live from the MGM Grand Convention Center in Las Vegas, Nevada, it's The Cube at Splunk.com 2014. Brought to you by headline sponsor, Splunk. Here are your hosts, Jeff Kelly and Jeff Frick. Hi, welcome back. Jeff Frick here at The Cube. We are at the MGM Grand, getting tired, Jeff. We're at the MGM Grand at Splunk.com 2014. We've been going wall-to-wall all day today, all day yesterday, getting a ton of great practitioners on. A lot of Splunk executives, analysts, press, smartest people we can find. We bring them onto The Cube, we ask them the questions to help them share their insight and knowledge with you, our audience. And we're really excited about our next segment and we'll be co-hosted by my co-host here, Jeff Kelly. I'm Jeff Kelly with Wikibon, and yeah, we've got another smart guest on Sanjay Paul, who's the VP of Advanced Services at Cisco. Sanjay, welcome to The Cube. Thank you, it's a pleasure to be here. So, one of the things we keep hearing here at the show is all around industrial internet, internet of things. Cisco, I think, calls the internet of everything. Talk a little bit about the role of not just Cisco, but kind of networking generally, and where that fits into the internet of things trend, if you will. Absolutely, Jeff. Thank you for the question. So, when we think of just internet, the way the internet is positioned today, we talk mostly in terms of the people connectivity. And I know you mentioned internet of things, internet of everything. The difference that Cisco's trying to make sure we get at is the fact that this new internet, or internet 2.0, is really a combination of people, as well as all the devices that are going to connect on the internet are already connected. So, the internet of things or internet of everything is really fundamentally this notion of pervasive connectivity between all of these components that will have addresses that we can connect with that will actually probably relay information back to some central location. Now, in terms of the role of the network itself, recognize that all of these devices will now communicate on this network. So, the network now watches for all of the traffic between these devices, between the people, and that makes for a lot of good analytical data that I can now harvest to look for security attacks, to look for potentially trends that might help me with respect to productivity gains, potentially some cost savings I can have. Where are inefficiencies in the system? So, in general, what is happening now is where we see the play for Cisco in the internet of everything is this notion of, I can look at all this traffic and once I ingest that traffic into some big data analytics engines, now I can harvest that for some significant gains that I would never have had before. So, for us, that's the benefit of where we can... And where do you envision the collection and analysis of that data when it's a bunch of connected devices that are associated with different entities, different companies, different workflows? Absolutely, so first of all, in many cases, we actually want to ingest that data into engines like what Splunk has, for example, so we can quickly harvest that data. That's a very important aspect. In some specific use cases, Jeff, we want to be able to harvest data literally in real time. I'll give you an example. In the case of security attacks, I want to be able to know that an anomalous pattern of network behavior was happening at a certain point in time and I want to be able to pick that out very quickly. So it's important to kind of, almost in some instances, monitor and track that behavior at the edge device itself, where this information is coming onto the network. So we have products that will do what is called fog computing, but probably more commonly called as edge computing. The reason we call it fog computing is it's like a parallel to cloud computing, saying there is a cloud computing and then there is the fog computing in the ground. But fundamentally, the notion here is we are looking for edge computing, some sort of analysis that can be done closer to where the action is happening, so I can quickly react to it. And in some instances, then I take some of that data back to a cloud location if you were and analyze it over there. So we're working with, in this instance, Splunk to make sure that we have a decentralized model of analysis, if you would. So security is one example and I know Jeff, you asked earlier about industrial Internet of Things. So when you think in terms of industrial Internet of Things, one of the important notions here is we may need real time analysis for the industrial information that we are ingesting because the feedback loop to make sure the devices are operating in the most optimal manner needs to happen in real time. So this is not a security use case, but it's more I'm actually providing a certain sort of feedback and there is some software that will now control these nodes in a certain way that will behave based on this feedback. And therefore, we get a chance to do some of this edge computing to kind of impact what will need to happen at the, let's say in a factory floor, if you were correct. So, okay, so when you're thinking about an industrial Internet use case, if it's a wind farm, if data is flowing to and from the different wind turbines, you want those real time capabilities to maybe adjust to new wind patterns or even maybe demand and you're going to kind of automate that process using intelligence, whereas you could take that data, collectively bring it to the cloud and you're going to maybe do some more historical analysis to look for trends, build models and maybe you'll feed into the real time systems. But you've kind of got those two different layers and there's probably even a middle layer there too where you might be, you know, the edge device, you might have the whole wind farm, maybe you'll have a kind of a central location there and then off to the cloud, so. But from an architectural perspective, in terms of networking, now, so is it about using software to do the analytics on the edge devices or is the analytics built into the device itself? In other words, is it a high-end smart piece of equipment from Cisco or is it Cisco, is it moving in this commodity hardware area where you're using software to add the intelligence? Right, not a very good question. So it's actually a very insightful question. Let me tell you why. First of all, let's go back to the wind farm example. So you have a single wind turbine. It has a microcontroller on it. So fundamentally, if all I had was one wind turbine and I was communicating with it from the network and saying you have to run at this speed based on this wind flow, whatever, so that's one level of communication. So based on this very simple feedback loop with that one single device, I can make it behave in a certain way. Now, if I add more, let's say a hundred wind turbines together sitting in a farm. Now, interestingly, especially that example is very good because one wind turbine's performance actually affects the performance of the other wind turbines. So it's almost like when you power these up, you have to power them up in a certain sequence and when you power them down, you have to power them down in a certain sequence. And therefore, the whole hundred wind turbine system has to act in a certain way to be most optimal. And there is a software component that sits on top of the hundred wind turbines where now we can monitor the hundred and then make sure they're all optimally efficient as a system. So, and then you take it back to saying over a historical period, how did this behave and what can I learn from it? So that I can impact the overall behavior of this system. Now, this is where I truly believe the sort of the evolution of the industrial internet ultimately will go. The hundred wind turbine example is more a real time example right now, but even other systems, factory floors, supply chains. Actually, I would say the best example of an industrial internet today that most of us can even relate to is the autonomous car, right? So there you have a network of nodes, peer-to-peer nodes talking to each other in real time ingesting data. And when they ingest the data, it reacts to the context of that data and figures out what it needs to do differently. That's the best example that we can all immediately say that is what the real industrial internet impact is going to be. However, today it's more evolutionary steps. Today it's more I can connect to the engine, I can get feedback from the engine, I can connect to one wind turbine, I can get feedback from there. As we go forward, you will see the opportunity to create sort of these systems that will then act as a homogenous system. Because we're in Las Vegas, I should say at least one more example. The Bellagio Fountain interestingly is also an example of a peer-to-peer system where the water jets, the color of the light, so the lights themselves and the music are all orchestrated and in real time, they communicate with each other to make sure they are all synchronized. It's not pre-programmed. See, I just assumed it was all pre-programmed. I thought it was just running off a script, but let's take the wind turbine farm another step further, because it seems to me that one of the hurdles that'll have to be overcome is really all these connected systems, right? Because you've got the ecosystem of the 100 turbine wind farm, but that's connecting to a power grid that's connecting to, say, a demand mechanism that's connecting back to three or four different types of distribution points on that electricity based on market forces. So it seems like it's like the API economy on steroids because there's got to be a lot of efficiency and now there's conditional feedback where if anything breaks along the chain, it could be completely out of my control whether my turbines are running or not because this something broke down or didn't give me the right feedback on my missing an opportunity to send power to municipality A. So do the feedback loops and the two-way communication in the network today work that well? Are we developing new stuff? What are you guys doing to kind of address that machine to machine feedback loop with some level of intelligence and or drawing intelligence off of some other brain if you will? Yeah, so Jeff, you've hit on a very important characteristic that we have to make sure in this industrial, internal things world the network has to literally have from day one which is resiliency, right? And resiliency is kind of tied also to security but fundamentally, no matter what it is, what event might happen, it could be a security attack, it could be a fact that some link broke down in the network. So whereas in the world of internet of our current internet, we are okay with availability being sort of potentially secondary and confidentiality for us is actually primary. In the world of the industrial, internal things availability and resiliency is primary. Confidentiality I'm not as concerned about it's just devices talking to other devices. So the way we architect these networks in that instance is making sure that there are several redundant paths between these components so that if one or two were to fail that I can still make the connection happen because it's absolutely paramount. We never ever fail, you know, break down this entire chain of the different communication. But he also brought up an interesting point which Jeff had also asked about was this wind turbine form example, we talked about it as a system at the 100 wind turbines. And you pointed out, you know, take it back all the way down to demand response and energy is a very good example of a very real time response that you must have because as the demand shifts, you need to make sure, you know, your response from the utility grid itself can react to it. And that's why the whole supply chain can be part of that whole ecosystem of saying I'm going to react based on the demand and can be most optimal. But I just want to make sure we emphasize that. Yeah, and I do think the security is a concern. At least it just feels like that I'm just introducing a whole bunch of doors for people to potentially have a tax. To Jeff's point, how smart are those doors and where's that intelligence? And obviously economically, you can't put too much intelligence out at the edge. It's got to be in a little bit. But I would imagine that's got to be another real issue for CIOs at looking at these opportunities to think, man, oh man, am I just completely opening myself up? Or am I wrong? Everybody's already open because we're walking around these things anyway. No, so it's such an important parameter to consider in the design of any network that is tied to the industrial network, right? Because I know you're looking at your smartphone and saying, OK, I'm already connected. And yes, I'm vulnerable. For an industrial network, that is unacceptable. Because you're not talking about loss of cash or money. You're actually talking about potentially physical damage or potentially loss of human life. Sure, like with your car example, right? Somebody gets in and hacks your health care examples. Or the health care examples, yeah. So those are things. So for us, we put security as almost integral in the first phase of the design itself. And that means things such as we trust nobody. We trust not a thing that's connecting in. Every time you verify the connection, and there's always constant handshakes to make sure that that entire communication is in fact valid. So there are a lot more security sort of constraints that we put around these communications that we did not put necessarily in our current internet. So let's shift gears a little bit. We're at the Splunk show. So you're here as a customer, right? You're not here as a partner per se. So what do you think about the data philosophy that Splunk has adopted in their applications and how you guys can use that? So for us, the ability to take all this varied amount of information and plop it into these engines that Splunk has and start to make correlations between these. And in the world of internet of everything, it is almost very, very critical that we ought to be able to monitor the infrastructure, look for these security attacks, look for resiliency, look for something that may point to a potential problem, maybe not right now, but probably six hours from now. I want to be able to find that out sooner because the effects of this are very catastrophic. So for us, it's almost like these two things go hand in hand. I have to make sure that all of this data is ingested and it performs a network that's performing as it's supposed to perform. It's a very important collaboration. Have you implemented or have you already integrated into your other management tools within the networks? Sweet, how's that working for the early days? So today we have implemented Splunk more in our IT operations. And with some of our customer environments, we are in the process of evaluating the implementation of the Splunk platform itself in some of our Internet of Things environment at the moment, yes. Okay. Yeah, and just to kind of expand on Splunk, I mean, just what you're taking on the company and Jeff talked about their approach, but they clearly, Splunk has an ambitious agenda. They talk about machine data and clearly that's what we're talking about in terms of some of the use cases, but they talk about analytics for everybody. They want to expand beyond just, they don't want to be pitching holes into just some IT monitoring application. And it seems to me that they're starting to encroach or they're going to increasingly encroach on the more traditional BI guys who have been always been struggling with getting not just adoption, but that more real time view. What's your take on Splunk's prospects as they kind of expand beyond kind of the coordinates where they started and becoming a much more general use platform? Yeah, so very good question, Jeff. And I think fundamentally the first Splunk is literally ingesting all this data and as much data as it can get, the better it is because now you've got the base foundation. The next question is then how does it continue to evolve and grow? So then that's where your point about the BI or the business analytics or business intelligence. So fundamentally then I take that data and I start to make sense of it. And interestingly from a customer standpoint, for me that is really what is the most critical thing that I have been looking for. It's good you're telling me about my log files and you know, IT operations, that's great, but that thing just scratches the surface. So from a prospects perspective, I think Splunk has tremendous prospects going forward if it starts to execute on this vision of taking this base of data and over time building upon it and providing the capability to analyze this information in a very simple manner without having to learn some complicated query languages. Plus on top of it, the second part of it is to visualize this information in a manner that makes sense to me as a business owner. So to the extent that we can actually have apps that say I'm a marketing business owner, tell me, you know, for my application, what am I, I may be an HR business owner, I may be a finance business owner. So based on this data, I should be able to use some very simple, you know, graphical queries and get the information I'm looking for. So the prospects for my standpoint are tremendous because I think this is the only company with its time to market advantage and as it's ingesting more data, it has a lot of potential going forward. Yeah, it was a really interesting point and I think, you know, one of the issues with when you're building a general purpose platform that can apply to a lot of different use cases, the challenge there is, well, okay, I don't necessarily have the business domain expertise, they being Splunk. But what they're doing and I think is really smart they've fostered this community of users of their customers and they've opened the platform up through SDKs and other methods to allow their customers to start building applications and customize. So, you know, Splunk can maybe get you 80% of the way and then let their customers kind of customize for those particular use cases because those are the people that have the domain knowledge. They know if it's a marketing person in a retail environment, they know what's important to them and what they need to look at and they're going to learn over time what are the best practices in terms of visualization and Splunk can learn from that themselves, and make that available to other customers, I think the sky's the limit. Yeah, I agree. And where they sit, in this larger world of big data that we talk about as an industry in a buzzword, they're one of the few companies that are out there delivering applications that are creating business value. A lot of that market right now is around the lower level infrastructure of the plumbing around Hadoop and some of the SQL on Hadoop capabilities. Splunk's one of the few companies you can go to and they can say, well, here's thousands of customers and here's what they're doing in industries from transportation, to healthcare, to the security analytics, to financial services. So I would say they're in a really good position. Now let's take the flip side. I mean, what do you think of the challenges for Splunk as they grow? Yeah, so I think, so first of all, the challenges as long as I sell to a CIO, what happens with that is the CIOs are very technologically savvy and they're very open to trying out perhaps like an open source implementation. In fact, we're doing that at Cisco. Rebecca Jacoby has decided to try the Elk implementation because she believes there may be lower costs, you know, whatever. And for a CIO, that's always going to be a challenge. You know, there's going to be like, why aren't you doing open source? Because typically their followers are largely the technologically sort of savvy folks that, you know, and but for a business owner, that's different. Business owners don't care as much for the technology behind the scenes. They're looking more for, give me the output. And once I get that kind of output from you, I will say, you know, I will spend my money to get the same type of information or more in-depth information. So they're more looking at it from a business, you know, result standpoint or an outcome standpoint. So to me, you know, Splunk's challenge, especially with the open source movement, et cetera, is going to be having to shift the buyer from a CIO buyer to a LOB or a line of business buyer. And making sure they're more relevant to this new buyer because that stickiness is phenomenal. And nobody can take that away. And as I said, the LOB buyer doesn't care about the underlying technology. And that's the challenge that they have. Yeah, that's a really good, very astute point. I mean, the business owner does not even probably know what Hadoop is. It doesn't care. They want to know, how are you going to solve my business problem? As you say, they are willing to pay if you can deliver a solution, whether it's open source, not open source, whether it's, you know, big data or not. If it's delivering value, they're willing to pay. So very astute point. Well, Sanjay, thanks for stopping by. I'll give you the last word before we're getting the hook. If you had to put a bumper sticker on what you've seen here the last two days, back of the cars as you're driving back to the Bay Area, what would it say? It's all about business analytics. It's all about business analytics. That's what I would say. All right, with that, we will wrap the segment. Again, you're watching theCUBE. We're at the splunk.conf event, fifth annual DNGM Grand Las Vegas. I'm Jeff Frick with Jeff Kelly. You're watching theCUBE. We'll be right back on our next segment after this short break.