So, in summary, this whole process of predictive analytics, combined statistics, data mining, extracting good data, sizing it, manipulating it, cleaning it, turning it into smart data, and machine learning. Picking a machine learning algorithm and there's many of them out there. Extracting good data, what we care about, properly preparing it and use machine learning algorithms. It's time-intensive. I can attest to that. I spent several weeks working on it just to get a six weeks of data and get the analysis to where it was. Failure is more common than success, you've heard that before. I want to just reiterate that. Authors of a couple of the books that I read were really adamant about that, so I wanted to share that with you. Again, don't just think you're going to take a bunch of data and throw it at a machine learning algorithm and these wonderful answers are going to pop out. It takes a lot of work. This is a short video from the Bosch ConnectedWorld Conference. It's 16 minutes long, so we're just about done. I got a couple of slides after that. I like this and I wanted to share it with you. The topic I'd like to talk about is some basic principles as I personally see it as well as my company and many of our partners and customers, and then also what we're doing to address these basic principles. Now, from a perspective of a conviction I have, if you don't have some basic principles to hold onto in an innovative world, in a world that's emerging, you can be swayed by anybody's opinion. You're in a meeting and this guy says that, or this girl says this, and it's easy to be swayed and get off track. So, you have to have some fundamental philosophical basis. We hold to these principles that we established about a year and a half ago and are building products and services around them. So far, it has been very successful and actually delightful to see how this is taking off. So, I'm happy to share with you at this great forum sponsored by the Bosch company. I don't quite know how to switch over. Okay. Here, I like talking about the seven principles of the IoT, and then I want to give you two other sevens. So, there'll be seven principles, they'll be another seven things and another seven things. All packed in to 16 minutes. Therefore, you must take notes because there will be a test at the end of this session in the spirit of educating here, and I hope entertaining as well. So, please let me move on to principle number one. It's a statement and an assertion. The IoT presents a big analog data problem. Now, what is big analog data? Well, it's big data. I think you've heard big data, it has achieved celebrity status. But it's of the analog nature. Most every thing in the T, in the IoT of things, is pent-up analog phenomenon. Such as vibration, such as light, such as sounds, such as location, acceleration, particulate moisture, speed, heat, and temperature. For example, all these analog phenomena are digitized. It's important to know that it's really, really big. Some of my colleagues generate petabytes a day, some generate terabytes in just a few seconds. That's how big this is. So, this is very profound. If you're dealing with big data, if you like big data, you're going to like the industry of IoT and the future of IoT, as homes and businesses, and cities and vehicles are producing this data. So, big analog data is a term that has been trademarked frankly by the National Instruments company, my former employer, and it represents what is sourced from nature, people, and things. It is the data from the things. Okay. Let's move on quickly to principle number two. The IoT offers perpetual connectivity. What would you do if you were perpetually connected to your products, or perpetually connected to your environment, or perpetually connected to your customer? The Apple company is perpetually connected to me, which is why they tried to convince me to buy the Apple watch. So, there's a sell-up notion of perpetual connectivity, you can monetize that. One day I woke up and I turned on my iPhone and the app for the watch was there, including an advertisement. What was the inference there? Well, you already have the app you might as well by the watch. That was the inference here. So, that notion of perpetual connectivity, one idea is to sell up. But there are really three M's, and that was one called monetize. The first M is to monitor. If you can understand the behavior of your customers and your products, there's a lot of value in that. Next is to maintain and manipulate if you will to add another M. Flashes upgrades, the Tesla car has that at night, it gets upgraded with new software, et cetera. Then, of course, monetize. We talked about that being able to sell up more products, services, and advertising. There are other benefits, but it's a nice way to look at it as you build your business case. Very quickly moving on IoT data is really real time. Many of us here in this audience from the IT world, at servers and workstations, and networks, and we think real time is when the data hits a network switch, or when it hits a nick inside one of the servers so it can be processed. Well, that's not when T0 starts in the IoT. T0 starts in the IoT way back at the things on stage one. Before it even hits IT equipment, there's a whole life where it's actually old then. So, I'm introducing now a four-stage template that is several years old, it's very durable. It's not perfect but it is a great way to talk about this amorphous thing called the IoT solution. Let me just describe it very quickly. On the left are the things, it's end-to-end. You need sensors that are either wired or wireless to actually capture data, and then you need a way to aggregate it and to do some upfront processing and maybe destination, or preprocessing or conditioning of the data. That's where the IoT gateway falls in and the data aggregation systems and back DAQ systems, which are tens of billions of dollar industry right now. Then it goes over to IT. This is where it hits IT for the first time. A workstation or a server out of the plant, or a manufacturing floor, computer system, an embedded system, out at an energy station or something like that. Or inside a vehicle. Then, of course, the infamous data center, or cloud. Well, I call the cloud as just a data center that nobody's supposed to nowhere it is. I think you would agree with me. So, there's nothing mysterious about that. So, the cloud and the datacenter are the same. So, there's this notion of sensor all the way to the cloud, which is very possible, but there's a lot of business, a lot of value, and a lot of action that happens in between the thing and the cloud in these stages. That's why this is so fundamental and one can impute visualization, as you can see here. Analytics, data flow, left or right in control flow equally important. What my colleague Andy will talk a lot about controlling things. Therefore, one can impute management, different stacks et cetera, across here. Some of the winners in IoT will be able to synthesize these and make them seamless for the end-user, or the customer. So, back to principle number three, this is real-time for IoT data, this is real-time for IT data. Those who understand the life cycle from here to here will win. How soon? Do you want to know your asset is going to catch on fire? Probably really immediately. How soon do you want to know if there's a little girl in front of your autonomous vehicle that you're driving at 40 miles an hour? It's a little girl, it's not a plastic garbage can. How soon do you want to know that to take the correct evasive action for the automobile? Because clearly, someone isn't going to sacrifice their car perhaps to avoid hitting a plastic can. But many of us would always agree that to avoid killing someone, we'll sacrifice the vehicle, take a different evasive action. So, do you really have time to go all the way to the cloud and back? That's a very strong phenomenon, a very strong principle in this world of cloud intoxication that we have. Personally, I've nothing against cloud. I would assert my company has the best cloud offerings in the industry, with our partners General Electric, with our partners Microsoft Azure, et cetera. But the point is there's a big business for doing work here and getting answers here as many in that industry would say. So, let's move on to number four. The insight can, therefore, related to what I just said: "the gained in a multiplicity cost a spectrum of value." There are four domains of compute effectively, and the data insight can have real-time early life in motion at rest and archived. Many times the data is really at rest by the time it even hits a server, or workstation in this formal definition. Here's where it is in motion and in flow and out. This is not a perfect model, not worthy of debate necessarily. But it shows just the different point of view of data from the IoT versus the IT world. You know where I come from. Three types of insights that I like to look at. You can drive business insight, where's my inventory, how long is that line in the manufacturing the products, how long is a line at a retail store of customers waiting? Business insight obviously can be derived. But engineering insight, how soon will that turbine at a power plant fail? Or how soon will the robotic arm need maintenance in a manufacturing floor? That's engineering insight. Scientific insight, which I include medical insight. Is that tumor benign? Is that a new subatomic particle? Like CERN, the Large Hadron Collider in Geneva discovered a couple of years ago a new subatomic particle by connecting things like particles and control systems to a network, and they created what I would call the scientific internet of things to discover and promote advancement in technology. So, those three insights are key and then it's good to understand your customers and your business, what are you trying to achieve with respect to that. Not everything is a business insight, especially if it's a government lab or a learning institution that wants to use the IoT. Okay, let's move on to number five of seven principles. There's a trade-off. Four domains of compute. You can compute and analyze data at the sensor with smart sensors, but that's very rudimentary today. Contrary wise, you can move it over and do it at the IoT gateway. Most IoT gateways today are low-power Atom processors or ARM processors or maybe some proprietary switching and routing processors. So, they're limited and are generally closed in this dimension. You can go over to traditional IT and servers PC right at the edge, and then you can go up to the Cloud with of course, rows and rows of data farms and data centers to do deep, deep analytics. Therefore, if you want immediacy of your answer, if you want to get a temperature, if you want to do a moving average, or perhaps a fast Fourier transform or rolling average or something, you can get it immediately, but that's not very deep. If you want depth and, for example, you need to compare with 17 other sites around the world or seven years of data because there's litigation going on, you need depth, but it's not going to be fast. Now, this is interesting. Now, listen. This is a mutual exclusive objective. Our customers in the IoT world trading off immediacy with depth of insight. If you can solve a mutually exclusive objective, you can catapult your career, you can catapult your company and your customer success. I learned this when I was young and my boss came to me and said, "Make a laptop computer as a lead engineer that battery lasts all day and it's light enough to carry around." That was tough 25 years ago, because why? Because I can make it quite heavy and it would last a long time, or I could make it very light and the battery lasts 20 minutes. So, I told my boss, "Well, that's a mutually exclusive objective. I can do one another." He said, "Well, that's why we hired you, is to solve a mutually exclusive objective." Twenty-five years later, I'm sharing that lesson with you. When you can figure out how to get deep insight fast, you will win in many cases. I'm going to share some more very quickly about that. So, there's the trade-off and now number- let me move on here in initial time. Visibility is an XV. If you ever study big data, you learn that visibility that the Vs on the left, variety volume, velocity, value, we're adding another V, visibility. Why were the edges remote? You need to visualize data, you need to visualize application stacks, which are going to be more and more complicated. So, graphics technologies, remote desktop technologies like VMware Horizon, Citrix, for example. These technologies that are in the data center and used for PCs in remote workstations will be imputed at the edge. So, you can see there's a lot of predictive dimensions with these principles, and not only just statements, but they're predicting what will happen. What's the best way to predict the future? Is to build it, right? Who said that? No time to wait. I got to keep moving here. The best way to predict the future is to invent it, as well. So, we're asserting that and we're inventing a future here with some of these predictions. Moving on to number seven. This is the essence of my personal business and a lot of what my company is doing with my partners here. Data center class compute, which is normally reserved for inside the data center Cloud will shift out to the edge, that is pretty profound. What am I saying here? Well, lots and lots of high-performance server class scores that do deep computing, high-performance computing, enterprise class manageability, data center level virtualization and manageability, scalable storage, all the stuff that's reserved today in the cushy environment, air-conditioned environment of the data center will move to the hostile edge where it must be temperature, shock, vibration, heart. That's what we mean by shifting left. Shifting left, of course, from this four-stage diagram, high-performance computing virtualization containers are vogue today in the IT world. All this IT stuff is moving out to the edge in this OT, operational technology world. So, the combination of IT and OT is fusing and converging together, and that's what principle number seven is all about. So, as I move on here, why would you shift left? There are seven reasons. So, take a picture because this will be on the test. It's an open book test. So, you can use your notes. Take a picture. Latency, bandwidth, cost, security. If you go to the Cloud, it's going to expose you for security. Reliability, corruption, compliance and data sovereignty, there are some times when data cannot move away from the cloud. Okay, the answer is model today an IoT solution, and I have built these personally and been out there. Looks like this, lots of boxes together. We can converge them with the HPE Edgeline system. This is a brand new product category that takes several things such as on the left, high-performance computing, data acquisition controls, storage, manageability, remote manageability that's common in a data center, build it all into a single box. In open industry standards, this will disrupt the nature of our discussion today, the industry, because the industry is low performing. The industry is unconverged, the industry is closed, there are not open architectures. This is a new way of doing things at the IoT Edgeline. Have you ever built a stereo system like that on the left? You have to be a certain age if you want to have built that, and then you can't find the cable, and then you're swearing and it's 3.00 in the morning. I don't know if you remember those days. Okay, what did the Bosch company and others do? They converged on a single box. You sit on your counter, or your bedroom and you turn it on, it's done. Similarly, the Android people and Steve Jobs at Apple said, "Why don't I take all these things that you carry around separately and put them on a single platform?" That's the value of convergence. Computing at the edge to get deep insights fast and do it with a converged system. Seven reasons to converge. Think about your smart phone if you have an Android or iPhone. Seven reasons. There's less space, less energy, less latency, less cables, less deployment time, less to buy, and less siloing of the world of sensors and data acquisition and IT because you force them to come together. That's the value here. I don't know how many of you are members of IEEE and are signed up with them and get their emails. I get a lot of emails from IEEE, and most of them I delete, but oftentimes, it will be interesting article. So, this one caught my eye because it's very relevant to last week's in this week's material. This article pointed out this executive skepticism associated with machine learning and data analytics in this day-to-day transformation with data as the new currency, being the best provider of information. It's called the digital transformation. They performed a survey of executives. I don't remember how many executives that they interviewed, but it was some number. The survey came at one in five executives thinks it's a complete waste of time to spend any time looking at adopting any of this or even taking a look at it. That's 20 percent. You might say, "Well, 80 percent. They're down with that, or we're good with that, or they're willing to explore it." I would have expected that it would be at this point in time where we are now today, it would have been much, much lower like less than 10 percent, five percent or something like that. So, the article said they had a few points to make, and I just summarize them. They made these recommendations. They said, "Focus on business outcomes." That's important. You're not going to jump into this unless you have a business objective. You want to improve operational efficiency, you want to increase revenue, you want to decrease your costs, you want to increase the safety, or some tangible business outcome. Those things that I just said. Commit to execution. It takes a lot of work when you're just starting, and you have nothing, and you need to deploy all these sensors, and get all of these equipment in, and figure out what algorithms work, and how to extract all the data, and which data is right, and process all that data and prepare it, and extract good data and make it into smart data, it takes a lot of work. But the recommendation was to commit an execution. Sharpen your soft skills. I strongly encourage you to go to this link and read that technical skill stack. There's a link in there that links to another article called technical skill stack, and there's really good stuff in there for you, trust me. Go read that, and then put people first. Transformation results in changes to your workforce, and it's important to put people's lives and the employees of your company out in front and be thinking about them because some previous jobs would become obsolete, while new jobs get created. Every technological revolution that's happened, there were a group of people that were affected and displaced from their jobs and simultaneously created opportunity for different jobs and with different skills to come in and fill different roles as the business environment advanced and changed. Sounds to be all doom and gloom. I found this one from GE, and they make the GE predicts. This again hearkens back to this the power of one percent which we touched on earlier in this semester. These are efficiency gains. The smallest one percent can have a sizable benefits over 15 years when scaled up across economic system. So, this is savings in oil and gas, $90 billion in savings from reduced capital expenditures, $66 billion in fuel costs in the power industry, $63 billion from efficiency gain in the healthcare, $30 billion in fuel savings cost in aviation, and $27 billion in rail. So, it's a good thing despite the one in five executives thinking it's not a good idea.
