 Okay, good afternoon. Let's see. So this is how much time I have left, correct? So I have 40 minutes left. Okay, got it. Okay, first a little bit about me, I suppose. Which one goes next slide? There we go. Okay, a little bit about me first. So the first 13 slides are a little bit boring. I'm going to run through those and then we get interesting. So just to give you a summary of what this is about. So for the past five or six years, I've had one job only, which is within HPE and Microfocus is to fix big projects that have gone astray. And so it's been a global role. Indeed, that's not my real job, but apparently you need one of those people in anybody's organization, so that's been my role. So it's taken me sort of a lifetime of effort to get to that point where I can show up anywhere around the planet and sort of fix things. So we'll run through some scenarios, some factories in a smart city. But that, first off, a little bit about me. So I'm a big data and knowledge domain engineer who helps with smart cities and big sensor projects globally. So in that, I've done some smart cars, some smart factories, and even some smart space projects. So I've been around the open groups since 2001. It was the first time I rolled into the door. I was actually at Micro at IBM. My boss made me show up to that. And so I've been doing it sort of ever since. So I've done a few boards. You can hear about the biotope. If you're here at the open group, you know that we're involved in some smart cities in Brussels, Lyon, Helsinki, and Peterborough. Explorico, which is a Norwegian high-tech incubator, and ArchMission.org. We do space and orbital projects. We're trying to, one more project. We're trying to build houses on Mars. So I get to involved in some pretty interesting things. I'm the chief VA for Vertica, which is a big data platform that supports enterprise and IoT data aggregation and analytics. I'm an imber riddle guy, so obviously a pilot in the aerospace industry. I did my graduate work at Wash University and raised seven kids. I live in Seattle. All my kids are millennials, which sort of matters, I suppose, for the next couple of slides. So that's a tunnel boring machine. So I travel for my work, and I'm responsible for doing solution oversight and latest trends in drivers that our biggest customers might care about. And so I'm supposed to make big organizations happy and successful with our projects. But essentially, my oldest kid of the seven is a civil engineer. So that's a tunnel boring machine. That's a 10-meter bore. You know, there's not much human assistance that happens once it gets going. There are sensors over the cutting edge of it, and essentially there's a wealth of 3D data material, of the material being excavated. So data is wired to some data storage to show progress, to improve machine maintenance and design. So first of May, I was taken off for a project in South Africa. It was a mining project, and he called and said, Dad, we have all kinds of data from doing this work. So he's got his masters and his PE and so on. He says, anyone can drill a hole. Elon Musk is drilling holes, in fact. He's building autonomous machines. He's not hiring civil and machine and mechanical engineers. He's doing it with software engineers. And he says, I wish there had been a data scientist. So I pause first. I kind of says, well, why? He says, well, I want to be able to correlate and forecast what we can do with all this data, because right now we don't know what to do with it. We hand it to data scientists, and we get tired of them, and we let them go after three months because they don't have the context of what happens in the geotech industry, inside of a mine or inside of a tunnel bore. So how can we repurpose and sell it? So that said, we have a massive IoT and double IoT gap. This is an opportunity there. So IoT is all about data that we couldn't capture or stream in the past. If you could have gotten onto the table and done the weather, well, then we probably would have put you underneath there a long time ago. So we put a sensor underneath there. We put sensors in places that you don't want to go or be that aren't safe. And so today we could do that. It's been commoditized, and now we have all kinds of data. So we can collect it. We can use analytics. We can predict it, prescribe it. We'll find the future state what we should do when we get there, and so on. So we collect all this data. What are we going to do with it? How do we manage this insight? So executives are thinking, hey, great. This is awesome. You have all this data. I want to know how to sell minerals or coal as I'm digging it. So we have gaps in sensors in the field. We need a lot more of these sensors. So I hear this number being patched around about 20 billion sensors by 2020. And in the oil and gas business, they'll say about 40.8 billion sensors will be around until by 2020. That's about a half a trillion dollar business globally. So we have a gap in sensors in the fields. We don't have enough of them. We need a lot more. We need to understand how to use the missing data to inform the business fast. And there's new roles to require. There's capabilities that are missing, and we have a people gap. And I would say that's our biggest gap. So who are these people who are going to do the work to collect prep and do this AI that Michael's been talking about to deliver value to the business and global market. So the capability gap is about data and people. It's not about BI. It's the hard industries and around the world. It's data to value. It's mining manufacturers and natural resources like water and gas in sunlight and be able to bring that data, that sensor data up to business value and make sense of it. So if you had all the IoT data and the people, what would you do with it? What's the opportunity? And so mindset and culture. The first thing you have to recognize that we've moved from digitization to digitalization and the fourth industrial revolution. So I look around the room and I won't expect you to raise your hand, but if you don't have pretty clear in your mind what the four design principles are of Industry 4.0, that tells me that you're probably not from the OT space and you're in the IT space. Because in the OT space, we live with it all the time. Every day, every pattern, every project we think about or are told about those design patterns. So we've got a gap in how we think about things. So data patterns are the requirements for agile, fast, successful design. Blueprints based on experience and academics are in fact a thing of the past. Data informs design today. There's no more design informs data. So you've got to see data versus a fabric, a 3D mesh. And you have to understand how to extract value quickly, cheaply and simply. The workforce is changing and it's got to be accepted as the only way forward as new generation is here. And so, you know, us old folks have given the new folks a problem and now we're going to have to figure out how we're going to fix this. So we have this gap between OT and IT. So building a data-driven world is forcing historical analogs in terms of the crisis. We've gone from horses to cars, typewriters and software factories to robots and now IoT measures what's missing. So now collect and prep it, correlate injected with value, model business strategy from global events and drivers and data and now data is an asset. So as we move forward, we've got to think about who's going to be performing this work for this new paradigm. So it's you, me and the millennials. I hear a lot of bad talk around the world about millennials. I'll say a few things about millennials. First off, we don't realize the volume. In South Africa, there's 15 million millennials. In the United States, there's 68 million millennials. In China, there's 440 million millennials. That's more than the population in the United States and Canada. Data and knowledge workers are critical to our foundation. We have to grow. We've got to constantly be making our children and coworkers aware of what's happening with data as we move into this fourth revolution, this digital age. If you're thinking about retiring at eight, going from working from 18 to 65, you're mistaken because people are living to be 120. You can't just get to 65, go on the dole and expect that you're going to arrive for the next 60 years. We're all going to be working this together to make sense of all this data and where we're going with it. So I'm going to give you three short stories and then I'm going to jump into the standard fair that we do at the Open Group to show you pictures and graphs. So these three stories will sort of limp at the colorful slides and then we'll get into the boring stuff. So I get projects that are broken and here's an example of three of them that I think kind of interesting over the past four years. So about 30,000 kids die in Kenya on the Horn of Africa from viral water. It's a disease called schistosomiosis. So the big consulting in the banks had gone into the government of Kenya and said we can provide a drill network that will solve this problem three to five years, putting significant debt on the citizens. So HP at the time, HPE, we were asked if we knew of a way to support a less costly infrastructure to provide time to value something fast so we set up a focus group. And there was a young 28-year-old engineer who said, you know it rains the most of the time in the springtime there. They have these things called long rains. They're 200 to 500 millimeters of water. And he said quite quickly, why don't we just pick up the water? Why drill for it when you can pick it up? So the solution was credit with starting a wave of startups, emergency management records, data telecoms saving some thousands lives the first year. What did we do? Well, because I'm an aerospace guy, one of my peers in college is now the chief arc at one of the data science lives with Airbus. And so I was able to get a flight of satellites, a constellation of three satellites to fly over Kenya, to provide us 11 passes of satellite aperture radar data. With those strip maps, we were able to download that into some HP pods at the time, which essentially little containerized data centers. With that we applied a GIS solution to that. We worked with a vendor that provided truck mounted solar operated filtering systems that would pick up water. So essentially we had to come up with a scheduling system so we could show the drivers where the water was. We worked with some NGOs to show people where the water was to make rally points. We helped to establish electronic medical rally points. Water was cleaned up and picked up and essentially people got healthier. So we predicted and created accurate data models. Data on population health movement, water recovery were created and provided those to the government. So this startup turned into a little bit of an industry. And now that practice I am told is actually proliferating. So that's the first one. Second was Nepal. What happened there? Oops. What happened there was that we had a failed telecom project. So I was out to come and have a look. And essentially what happened is that we did the same thing that I suppose that everyone would have done. Would have gone in and answered the questions for what data scientists wanted to know, collect a bunch of data, provide the answers to it. And the CEO decided that wasn't good enough. So essentially what happened was that CEO said he wasn't going to pay. So that makes sense. And so it doesn't pay, you react very quickly. So what I found is that we did just what everyone else did. We didn't sort of pay much attention to what was happening at the control plane within this telecom scenario. We paid attention to IT people, to data scientists in the business. And so essentially what we did is I went in and I built a team and we interviewed folks that were in the control plane, found out what they were doing, they were collecting, provided that to IT people and to data scientists and now provided a new set of questions that have delivered the answers to questions that they had at the strategic level based on global events and drivers from data that has been collected in the factory, if you will, of telecom space. The next one, I think I'm going backwards in slides, that's what's happening. I'm doing two at once. Oops, one more slide. So the next one that happened was, and my slides are out of order here. I'm going to do that with a slide. So the next one happened was with the Nepal project. So around this time there was an earthquake. 27 April 2015, 1832 people were killed, 728 magnitude earthquake. And so to help our story along, CEO was rather upset. He was actually trapped in a building for three days. And so I worked with a Swedish company, Swedish researchers. And what we did is that while we were fixing this data problem, we also collected all the cell phone location data and realized that we had some 300 people were killed actually looking for people who were trapped in buildings. So we started collecting all the location data for where all the cell phones were. And from that we created some models and essentially even if there's an earthquake we know where your cell phone is. We know the pattern of where you normally would be. And we provided that solution to folks. So New Zealand. So New Zealand, you're going to hear about that. And look at my slides whacked again. New Zealand, you're going to hear about that a couple of times. So essentially what happened there was we promised a safe city solution. We started out with some 700 traffic cameras. And a number of teams failed. And one of the reasons we failed is this goes to this data interoperability problem which you'll hear me talk about a couple of times. Data interoperability is this notion that they have 32 different camera models there. And with those 32 different camera models, what ultimately happened. I'm going to grab a glass of water. What ultimately happened with those 32 camera models is a problem that would happen if there's 18 people in the room here. I hope you don't mind me walking around. There's 18 people in the room here. And imagine what happens is if I gave you each $10,000 and I said, create an IoT gadget that gives you your date dime stamp, your first name, and the temperature, it produces that data every second. How many different data models would I have? Somebody yell out, there's 18 of you. 18. How would I do like to integrate 18 interfaces? You probably wouldn't like that. Do you want to pay for that? No, this client didn't either. So I had 700 traffic cameras by 32 manufacturers. And so by the end of the first Friday, I wasn't going to tolerate that either. So one standard became the way. So you can start off with the best standard of the room, but that's going to cost you time. So I started with a done the standard of the room. And ultimately, after a couple of weeks, all the cameras were working. And that turned into we're going to go to 7,000 cameras in northern New Zealand. And so I just got back from there a day before yesterday. So if I look a little bit weird, it's because I'm a little bit hairy from travel. But essentially, we did a number of projects there, but we'll go through some of those here in a few moments. So business to data driven. So what's happened with our little world in the last couple of years is the design has changed from tradition and authority and by academic leader to one where, imagine if you build a car with four wheels. The car with four wheels is not going to change, but the data will inform that design. It's not going to change the way that design is. And so business process is the same way. Business process is being challenged by data. So it becomes data driven design. And for the first time in history, you can't hide from how things are used because IoT gives you that missing data and you can see how it's actually doing. I actually know if you went home and watched YouTube for the weekend rather than going at work and working like a Trojan. So history and culture are traditional, typically informed policy in how we do things. We do it this way, this factory. We do it that way in this factory. IoT and big data and data driven changes how we do that in our cultures, communities, cities, business, and factories. So with that, now I'll jump into the more traditional stuff and I can put down my laptop and stop looking at it. It's good. So there's IoT and there's double IoT. And we often confuse that and I often hear IoT all the time. But IoT tends to be those things that are about personal stuff and IoT can deal with it pretty regularly because it looks like it's consumer oriented. And double IoT is that stuff that's more hardened and lives in the plant in the factory, if you will. And that definition that's up there is one that banded around an open platform three or four years ago and I've added a couple things to it. With capabilities increasingly seen as independent and autonomous, with an increasing demand for interoperability of other things often heterogeneous. The wires go away, folks. The BMW 2020 car doesn't have wires to break lights. It's all wireless. Things talk to each other. Design principle number four, Industry 4.0, means things talk to each other. Design principle number one is things are interoperable. And so that's the differences between the two. And so this is a list that I've kept up over the last couple of years of failing at projects because I do get to fail. I'm an iterative guy. I get into projects that are well over my head, experience and level, but they've failed so I get to fix them so I can go home. So I've learned a number of things. One is you've got to understand the OT IT culture, like my kid was saying. The reason we fire these IT guys that are data scientists is they don't know anything about geotech. Why would you? Michael, do you know anything about bulldozer parts? Why would you? Right? It's crazy. If you were in the aerospace business, why would you know what a mudball is? Does everyone in the room know what a mudball is? It's an asteroid that's got a bunch of water and mud into it. They're generally frozen. If you harvest them, you can take the water and go to Mars and have a glass of water. But not every data scientist knows that. So an aerospace engineer might know that. So this notion is we've got a variance between what we know. Reduce process variability and energy consumption. We talk about blockchain all the time. But how many people are roomed? Don't raise your hand. Know the cost of a blockchain transaction. Would you be surprised that the average is about $30 for transaction? Would you be further surprised that the PUE associated with a transaction can average $130? Well, it's because that's an abstraction. It's away from it. It's in the factory, if you will. So the idea of these observations are sort of my little list that you're welcome to take and steal from because essentially I've never had a creative thought in my life. I steal from everyone. But some of those topics look pretty interesting and might be things that you think about regularly. One, improve ability to attract a modern workforce. Having seven kids, I've seen everything as they've gotten out in the workforce. And I have a, well, it's tough to say when your kids are doing stuff, but I have a offspring that went into a well-known famous companies that you would think that young PhDs would want to have jobs there. And when she went to the interview, she happened to stop by the men's bathroom and then the women's bathroom and decided she wasn't going to work there because workers weren't treated fair. One of the things about millennials that I can tell you from experience is they're not going to tolerate that sort of childless childishness. We have to treat people fairly. And what you see in the OTI team merging is this notion about treating people fairly. In my world, I get to go into coal mines in South Africa and factories all over China. And in that space, I see lots of things. And as an American sort of traveling around globally, you sort of have a viewpoint about what you might think is fair. And that certainly plays out in my work with dealing with OT and IT. The notion of creating continuous improvement programs, and Michael has mentioned a little bit of that, is you're going to fail at things. You're going to come up with what you think are good AI or machine learning metrics. As you get there, you're going to, on your mode one, or your first phase one, it's not going to be very good in phase two and three. It'll continuously improve. That last one, number 11, is critical. Enterprise event correlation time to evaluate traceability, both top down and bottom up are critical to the story for the rest of this talk. Here's a few buzzwords. If you scan down through them, they're mostly kind of interesting. I'm going to pick up on a few. For this discussion, you should probably know what an M1, M2, that should be an M3 machine is. An M1 is a machine that's probably created before about 1985. It doesn't have much other than electromechanical sensors on it. M2s have a have modern day sensors put on them. And M3 are native today, industrial IoT, 2018 machines, essentially those robot things that you see with three or four axis arms. They're built in IoT. And we have to think about all those, because all those eventually are cut-deducted to some kind of an IT. So time to value HPC, which is health performance capacity, are things to think about. Time, latency, and interoperability, the three vectors that you most care about when it comes to big data. When you're moving from OT to IT, scaling is different. In IT, you don't typically think about petabytes, but it happens all the time in the OT world. We typically don't keep things for more than a shift or two. Dev Null is our friend. I hear real-time, I've heard it about six times the last couple of days. Real-time is not happening at about anyone you know, folks. Unless you're dealing with microOS, which is a real-time OS and a time-sensitive network, you're really talking about near real-time. And getting that perception sort of clear, I think, is important. And last on this page, I think I want to mention vocabulary taxonomy, reasoners. Michael, you called reasoners another word. But this is a notion of applying logic to your local ontology. That's an important concept for IT people to get. You're going to have words that are used in a factory, and you need to take car and automobile and build an equivalency out of that. That's going to take a reasoner to apply logic to car and automobile to understand that both of them are motor vehicles. Not necessarily a motor vehicle, they're just vehicles. They could be battery-driven. And to drive those to taxonomies. And Ron talks about that a lot in terms of what we can do with ODEP. This is a linearity. It's a pretty simple one. You'll see a sensor on the far left. This is an IIT sensor. We use Kafka a lot to stream up to about 1,200 streams of sensors from around the business ecosystem. Time series, typically, discrete mathematics is the most common thing that I do. A linear regression with three to five points of confidence intervals is the most common thing I do. Probably 90% of the time, that's the math I do. When I get to the control plane of the edge, I do curation because I can't send all that factory data to IT. They would smother with it. So what I do is I curate it. I moderate that data and essentially use some SQL or some kind of code to send a small portion of that data on a regular and periodic basis to IT in the business. And their geospatial, asset management, maturity, date time, vocabulary are sort of updated. We can enrich the data with weather data, what's happening in the business, and so on. And ultimately now present to the space of a business data warehouse where analytics contribute to business value in the presentation layer. Quick graphic here. It's about driving scale. The more data you have, the better prediction you have. It's about getting large populations of things. Everyone should take a couple of classes and statistics to understand if I talk about 20 days, there's very little I can say about months. If I have six days, there's very little I can say about the population of days. It's certainly not a population of days or weeks. And so on, understanding scale and value there is critical because the more data you have, the more machine learning accuracy you can get. That collection there of six algorithms are the most common ones that I use in my work, in the OT to IT world to describe assets, to predict them, and ultimately to prescribe what's happening with them. And some other time we can go through what they all mean, but I think probably logistic regression is the most important one. Increasingly for my work in the traffic world, support vector matrices, eigenvectors, eigenmatrices has been critical to my survival. Those three little graphics, essentially, you're gonna want it to data defines process or modifies as I said earlier. I like a lot of SQL because business people want time to value faster. Most business executives that I know can't write a word of Python, they can't understand what to do in a dup so I give it to them in SQL as quickly as I can. The more structure you can apply to a network, the faster time the value is. 60 to 80% of data's life cycle is involved in preparation, trying to homogenize it in some way to give it structure. That's a cost and a loss to the business. You should understand that. That should go away. These days I see about 12 data prep people per data scientist. That's horrible. It should be about one to one and hopefully we'll get to the world within Michael's lifetime. There will be no data practitioners. There will be spitting out data from sensors that looks like it comes from a common taxonomy by industry vertical. And the last is less movement. This is about data lakes. The idea is get data as close as you can to the context to where it was generated. If you move it across the edge, it's going to cost you a latency. It's also going to cost you because it's like we talked about the geotech folks. If you're not a data scientist that has an undergraduate degree in civil engineering and you're getting that data, you probably have a problem from day one. Leave the data where it sits. Apply metadata as close to you can to it and then to ship the value across the edge and to correlate the value. Some use cases for factories. So IT and OT, I claim, are merging. It's painful if that is for the people who are in the OPAP. It's happening today. Millennials, 38-year-old folks are showing up as CEOs. And what they're saying is, why do I want to have a department for my factory and another department for IT? And I'll tell you the truth. In 1998, I was working for IBM and I didn't know what IT people were because the portion of IT that I worked in was all mechanical, electrical engineers. We made robots. We had linear robots that go in these things called net store boxes and I thought the electricians brought us our machines because they wired up everything. They were wiring 220 and 440, 12 volt and 48 volts and handed us laptops. And so all of a sudden a couple of years later I learned about IT people. And so I come from a world. I've never worked for a CIO. I've always worked for the Vice President of Production or one of those outfits. And so I come from the OT space. And so in this world, in the OT world there's condition monitoring which is the this is what people do in OT day in, day out is about understanding the conditioning of machines to keep them up all the time. Preventive maintenance, prescriptive maintenance, using machine learning, a kind of AI to prescribe what's going to happen in the future and event correlation. Taking lots of events that are happening within the factory and more than one factory to figure out what we should do in the future. And I'll show you. This is a quick picture. This is like a weekly picture for me. This was in Taipei two weeks ago. You'll see some funny names of very common sensors. The i-Series sensors by Fanuk is probably the cheapest, the most common sensor that I see in China these days. And essentially this is for a CNC shop. They produce machines that are sawing and cutting and bluing and taping and shaping of materials. And so those CME machines, they get events called crashes where the machines can sometimes destroy themselves. And so we'll collect data for individual machines and push it into an edge server, little boxes that live on the edge. And they have some data store capability that will store data for a shift. So we're looking for essentially maintenance problems that can happen within a shift and fix them. And then we'll save off some events for a longer period of time. And then we're producing that data for the control plane. For people that get HMI, human machine interfaces, where they're reading off that data. And increasingly, we're curating that data to go off to IT in the business in order to help them to do business work. Forecasts what's happening with multiple factories, for instance. So that's a very common picture that you'll see. Now this gets a little bit more noisier because this is the work of every day of my life. You'll see a factory on the left that's doing a product movement around the factory. You'll see some axles that look like they have bearings on them with sensors on each one of them. It's nothing to go into a factory in China or tie along with a million sensors in them. They're typically reading numbers every portions of a second or a second. That's what they look like with discrete time series. So I'm taking that content on the left and then I have many kinds of machines in the factory. When you look to the right, you're seeing essentially a random distribution of events for different kinds of machines or series embedded with them. I also get involved with digital twins is how to model those in software so I can help prescribe or predict us in the future using machine learning. And so those are the two patterns. And frankly right there is 90% of what I do with AI in factory these days. A couple of quickies, not to dwell on them. Other than Michael, where it says challenge, the last sentence, I put that in there for you. How big is 1.5 petabytes of data? It's 90,000 large pizzas or 75 tons of pepperoni pizzas. The deal is that when you get involved with OTIT value, you get a monster bit of value. Whereas you're separately dealing with OT and IT, you just don't get as much. It's amazing how much... And so in this case, this particular project generated $14 million of value in a short period of time. It required 50% less data scientists than they thought they would need to get some 287% return on investment. That's the value is when you merge OT and IT together. This is a couple of organizations like this but essentially I get these little small factories like this and they don't have much shelving there. There's not much place for parts. They want just in time everything. So they want to forecast what's going to happen tomorrow, next week, next month, next year and then to build their cost models out for that. And so we can help them go in. I'm sorry, there's a name of a product in there in a company. Forget that. That's my mistake. But essentially I go in to help us model those kinds of solutions. So here's one with some kind of personal devices, personal IoT. Look at the size of numbers we're getting per day per month. 20 to 30 million individual training sessions a month, a billion data points. I'm commonly seeing record stores in the trillions now. That sounds crazy but we have to curate that and that data is coming from factories, it's coming from the field, being driven to IoT. So we have to curate that effectively. So understanding the context of that data so you can move it efficiently back to the business is critical. Let's get on a little bit of smart study content. I think everyone in the room has probably seen this picture. If you've not, this is Kerry Framling, Dr. Kerry Framling's picture from about, I guess it's been around now, five or six years now. But this is the biotope. And so Ron and every number of the other folks in the room here have been involved in this project for some years. This is five smart cities, Brussels, Lyon, Helsinki and Peterborough. And essentially it started off with things like emptying the trash more efficiently, dealing with crowdsourced lighting, smart buildings, having over pressure inside of buildings, dealing with some microgrid issues and smart meters, being more efficient or effective with motor controllers and air conditioners. And so we go to smart cities around Europe with this story. And so about four months ago I was in Brussels and we have 124 different parking lots there and helping them consider using ODEF with ODF and OMI for that implementation. So a couple of things I've learned. You probably won't see this in any books, but this is sort of my view of life. This is what it's like working for a big company. Executives take global events, drivers and data. They write some strategies, get some capabilities, hire some roles, human systems and machines, create some services and workers do some stuff. If you don't like it, you can go someplace else. The goal is sustainable profit, not about your happiness, folks. Let's look at this picture. This is a city. You flip that triangle upside down. Citizens are at the top because a citizen, a city is about making people happy. They don't like it. They'll vote out the leaders or move to another town or city. So citizens essentially vote for leaders based on policies of things the leaders providing, their services, roles, capability. It looks like that previous picture upside down. So that's something to think about as we do the differences between smart cities and our factories. They're imagining that governments can be run like businesses, but they can't. So this is my list, like the one you saw earlier, what citizens want. They want to be happy. They want to be safe. They want to have access to data and knowledge. They want to help and actually improve their community. They want to have a say. People move to cities I claim to get education and help the next generation, their kids. They want to be able to reject what they want. They want to be able to protest and be upset about things that they don't like. They want to use technology to request services. They want their city to work like their homes, schools and offices. They don't want a police state where cameras are watching the RV move, but they want to be safe. And so they'll take a lot of that. In Moscow, there's 450,000 cameras in the CCTV system. That's about 10,000 cameras in Northern Auckland. But people are not rebelling against it because they see value in safety. So people will trade this notion of being watched that they can see value that's being returned to them. And essentially that last sentence I think is important because it's one of the hallmarks of the digital era. People want to be able to participate fully in the city life to be able to access services. 54% of millennials don't want to own a car. They would rather have good public transportation. People don't want to be lonely. So safe cities. Safe cities is a subset of smart cities I've gotten involved in a few of those. This is one that had concepts like license plate reading, video processing, specialized cameras for finding out the kinds of things people are carrying over their shoulder around their waist, emergency 911, parking centers and so on. And so you can see that just looking the size of that not very many citizens there, about a million and a half, about 4,000 traffic cameras and so on. But I've gotten involved with that. And one of the things I've learned to be successful in this space, public meetings are required. Whoever was trained to be an EA or an engineer to do public meetings, I do. In fact, I've been to them when people were upset through pieces of bread at us. At least they don't throw tomatoes and shoes. But people want to say, and they expect when you go and especially you're fixing a thing, that you understand what's going on. They want to ask someone who knows what's going on. Citizens will not tolerate someone in a suit going out there and just telling them a story. They want to know the value. So this is a little AI story. Essentially what happens is that once we put in smart cameras and smart traffic lights and so on, I wanted to see some application of AI. Citizens are asking for this. So the problems I mentioned earlier, heterogeneous cameras, all these different data streams, they are asking questions like, when's the best time to ride the bus when it's raining or snowing, when it's afraid next week, and when I have trouble getting on the bus, how many people are going to be on the bus and so on. And new data sources and models, you start with the Goldilocks model as Michael mentioned, you want to add new models on top of it. How do you do this? And how do you do it prescriptively? So on the left you'll see, essentially these little AI stories that I do are three steps, collecting lots of data and data sources, hoping for heterogeneity. I have to spend a lot of time doing data prep to ingest prep and then process the data and ultimately visualize it to value. So here's a little one. This is about a two week little project to essentially collect 10 years worth of tickets, 10 years worth of weather data, current data upon weather, current ridership on buses from sensors that are on the bus, current sensors saying how many people are on buses, and ticket sales of buses today and for year to date. And with that we're able to predict 80% confidence, plus or minus one rider on the bus. And you can go to this particular city, I don't think I've said who it is, or I'll be guessing the picture. You can go to the website and you can actually tell how many people are on the bus based on this little application. So that's machine learning. So this is a conceptual architecture. I use it a lot. This was done in Russia for a company that has a Megan cell phone towers. And these conceptual architectures are great for EAs because you can essentially show value within a couple of hours. You've got Stratage all the way down to the sensor. And now let's apply an AI model onto it. So where those smart towers, smart cell phone towers are reporting data, we don't have the bandwidth to support all the data that they're getting, that they're gathering in the field from weather, vibration data, manious data, galvanic action, decay of the towers, angles and azimuth and elevation of the antennas, their GPS signals and so on. We don't have time or the space for all that. So what we do is we curate that data and send that on to the business so they can understand how performance capacity from the business. Now, I don't want you to see that, but I want you to see the shape of it because that's important in a minute. That's actually my implementation of an open platform. There are seven different vendors there, all of which don't care about that platform. But to have an open platform, it should look like a cloud, or excuse me, it should look like a service catalog. It should allow people to pick things that they want based on their role and there should be some complex things underneath it. Some of those are in cloud, some of them aren't. Some of them are in cloudlets. There's big data. There's smart traffic. There's parking sensors. There's emergency 911. There's different kinds of camera systems and all that, all behind an open platform. That said, one last little picture before we go on our next little area. And this is about reselling data. I now have clients based on a talk we did in Berlin a couple of years ago. We helped New York City sell their big data. And so this is a process model we came up with. It turns out it actually works. You're welcome to have it. So some conceptual architectures. Those boxes right now are what I do for a living. I start with executives who have strategy. I make sure it's connected to global events, market drivers, and my data. I make sure it's traceable using Togaf down to capabilities, down to human systems, machines, and sensors, down to service catalog and the platforms, to some microservices or full-time big services, to data stores which are essentially containers of data that comes up from down below. I use some industry-forwarded design principles to a data model for many sensors. And then I go back up the stack. I can start with the sensor vendors, start from the sensor and go up the stack and do the same thing. If you've not for me with capability modeling, this is a couple of samples right there. You can read this as your leisure. So between IT for IT and doing some capability modeling, this comes from 2007. This is Betsy Burton's worth with the Gardner. I do these like every month. The beauty of capability modeling is it has nothing to do with org charts. It has everything to do with what the work is you think you do at a company. This is open platform three for not for me with the model. You can go on the website and see it. But essentially, you're going to see global events and drivers strategy capability maps. And from there, you'll see human systems and machines and look under third platforms. You're going to see what Gardner was calling nexus of forces. We've sort of made it our thing by putting IoT in there and so on. But this is where your data stores and analytics are actually coming from at the IT layer. And beneath that's the open business data lake. And the beauty of that is is that you can have all your different data sources being addressed to a service. So sensors may be dumping data into HDFS or Wide Store Analytics Engineer or Call Store or some database or file system. And from there, we can address it up the stack. Someone said it earlier. I typically say IoT eats cloud. And that happens certainly in many areas like the Nordics, Asia and Africa where clouds are thousands of miles away. And so we do these locally. Processing is local. Loss of 60% of IoT, I said this earlier. It's important to remember this. Cost latency and interoperability. Interoperability is the biggest probably have. It's the thing that we least talk about. I mentioned this earlier, ODEF, O-M-I-N-O-D-E-F. It's a clever picture in that it indicates all the capabilities that those three standards can use. And a couple last pictures. This is the edge about the middle of this diagram. Sensors on the left go on to curation engines. There's essentially some kind of a data store or flat file that's collecting the sensor data and curating it. So we're not sending big data. We're sending right data or curated data to the edge to our enterprise data warehouse where we're enriching the data with other data sources using Kafka in fact to business value. I showed this picture earlier but just to get that essentially big data in the IoT world is three processes. Sensors to ingestion preparation process onto visualization. And that's the same model upside vertical but you'll see the control plane down at the bottom where OT is. You'll see the edge there. There's a bunch of different kinds of sensors and all those little words that you see that you heard the last couple days. You might not know what they are. I'll explain them if you ask me individually. And all that goes up the stack and you'll note that it goes up the strategy and you'll see smart products off the left where customers are and the data comes back down where services are and delivers back down to the edge. I find that a useful picture for executives that try to understand this fabric. Last slide. This is the Internet Industrial Consortium. I'm doing that at IIconsortium.org and there are 29 test beds there. The IIC is not a standards board. Essentially it's vendors who get together to show how things work with multiple products with multiple vendors. You have IBM, HP, Microsoft, Oracle, unlikely friends working together to show the value of a particular area and you go online and find those. And essentially by doing that they can help to show the value of their solution and there's an implied, there's a reference architecture that comes out of it. There's 14 verticals that come out of that space. And I'll mention on the far right I have asset to ontology, to reasoner, to vocabulary, to structure data and taxonomy and that's the order that you'll find it. Now I'm not spending any time on that today but this is a critical story and consistent time to value is the way we want to play this. So in that, that's kind of the story of smart cities and some factories of what I've been doing the last couple of years. So thanks very much.