Good morning! Good morning. We're going to get started here. Thank you for joining us in the business track. We just heard a great presentation by Citibank about changing the way they do business with MongoDB-- Substantial reduction of time and energy. We're going to switch gears here to the Internet of Things. We heard it really great presentation that Michael Olson this morning about how the Internet of Things is really the data that we've been managing.
And we have the good fortune this morning to have Dirk Slama here from Bosch. He's been spending years in their business development software group focused on how the connection of the Internet of Things can change the way we do business. He's a renowned speaker. He's worked with governments and cities about this topic. Really excited and pleased to have him here with us today. Before I turn over to him, please, if you haven't already found where you do your surveys-- at the end it would be great if you could give us your feedback. Each and every time you do feedback in that survey, you get entered into a drawing for a SourcePro tablet from Microsoft.
OK. Thank you very much. If you could turn off the volume of your phones. We've had a lot of interruption in the last group. And Dirk, thank you so much for sharing your vision on IoT today.
There's an actual revolution going on. I saw it with my own eyes. I came in on Sunday, I haven't been in New York for a while, and actually every single bar, every single restaurant, was showing the game USA versus Portugal. OK. So, I mean they still refer to it as European soccer. I'm not sure if the Brazilian people are too happy about this but-- so things are changing. Actually, I think this is a series of many revolutions. To stick with the analogy, I don't know if you saw the first game of-- I think it was France, who actually was awarded a goal by some technology that was used for the first time. It is actually-- well, it's kind of like an Internet of Things example. I really like it.
So it's technology that is basically used to find out whether the ball has crossed the line behind the goal, OK? So to avoid the Wimberley type of situation. There are multiple technologies out there. There are camera based technologies. There are also technologies out there where the chip is actually embedded into the ball, and then the magnetic field is used to find out actually whether the ball is in the goal or not. So, this might actually not be the biggest technical revolution. But in the soccer world, this was a huge cultural revolution. So the referee on the field, he gets a message on his watch and he can basically find out whether this was a goal or not. I think, actually, I'm from Germany, so there was a big discussion whether this type of technology should be introduced into the German soccer league on or not, and actually it was decided not to do this. So, like with every good revolution, this is going step by step.
There are a lot of people talking about the IoT revolution at the moment. So there's no lack of big numbers. So we did some research on our own. We predict that there will be about 14 billion devices out there by 2020. There's numbers from Cisco, from Gartner, there are people talking about 1.9 trillion, in terms of economic value add created by the IoT to the global economy. And at the moment, I think it's a little bit like self-fulfilling prophesy. So everybody's looking at these big numbers, at these big opportunities. Everybody's investing and a lot of things are happening.
Maybe another interesting number is actually a rather small number. So there was a statement from Gartner saying that by the year 2020, the costs of actually connecting, even the cheapest processor, will be below $1. And I think this is actually a really amazing number. If you think about this, that really means that we will have the business case to connect almost everything-- every device, and basically make a part of the internet.
And I think on the way to actually see this happening, we will actually see a number of revolutions. One of these revolutions is going to happen, I believe, in the telecommunications space. So if I look at the current global telecommunication networks, the average revenue per user for a tel co is about $200. The average projected revenue per device per year is estimated, at the moment, at $4-- that's about 2%.
Of course, we're talking much larger numbers. But the question is, will the current structures put in place by the carriers, from network management to billing, et cetera, be able to keep up with this? Or will we see completely new players entering this field. There are a lot of start up companies at the moment, like the [INAUDIBLE], et cetera, addressing the space. So I think this is going to be really interesting to see in the future-- how the creation of this infrastructure for the IoT is happening.
Let's take a quick look at why Bush is actually in the game. What is our motivation? Most likely, everybody in this room has heard of course, one way or another-- we actually get a lot of reactions, people saying, oh, you are the guys with the great refrigerators. They're really, really quite and great. I actually like this, but it's not all that we do. Bosch it's a multinational engineering conglomerate. We have about 300,000 employees worldwide. We are world market leader in car components manufacturing.
But we also do a lot of other things. There's a very good chance that most of the people in this room actually will carry some MEMS sensors embedded into their smart phones-- that you have in your pockets. That's another area for us. We had industrial businesses, drive and control technology, heating technology, et cetera, et cetera. So it's a highly diversified portfolio. And we often get the question, what are going to be the breakthrough things in the Internet of Things, right? And I think for the last couple of years, everybody was talking, actually, about the internet connected refrigerator. If you ask me personally, I don't see that happening in the next-- let's say 12 months or so. I might be surprised. But yeah, difficult business case.
Bosch also produces power tools. So some of you might have used our power tools for DIY type of work. And actually, I was always convinced that power tools is the second place where the Internet of Things is actually not going to happen in the foreseeable future. And actually, I was wrong. It is happening. And I will show you a great example later on in the example-- what we're doing with power tools in this area.
I work for Bosch software innovations. We are a 600 people business unit in the vast ocean of the 300,000 people company. And our task is basically to enable our customers and Bosch business units for IoT business models. Of course we're dealing with technology, a lot actually. But we're also dealing with business model development. And one of the things that we actually learned is the business model typically is not so much about connecting a thing. But it is typically much more about combining a thing based function with a back end service.
So to give you one example out of many, we've developed the so-called e-call service. It's a service that is actually, for example, now rolled out by Daimler, in than the Mercedes cars. And it's basically a sensor, attached to the air bag in the car. So if the airbag goes off, the sensor uses a GSM network to notify a call center in the back about this fact. And that basically tells the call center agent that most likely there's an emergency situation. So he can basically immediately direct emergency services to the suspected site of the accident. And you can imagine, time is critical and this type of situation.
So the real value actually comes not out of the sensor being connected to the airbag, but really out off the back end service that is really dealing with the situation. And this actually goes further. So the next thing typically is, once the initial situation is cleared, that you need to direct a repair car to the sites, arrange transportation, deal with the insurance, et cetera.
So our formula for success in the Internet of Things really has four ingredients. Of course, there's things, and the users of things, like the driver of the car in the example I just mentioned. Then there's the enterprise services. And last, but not least, there's the partners involved. So in this example, for example, the emergency services, the insurance, et cetera. Here's an example for how this business model works. Remote condition monitoring-- this is actually not one of these more futuristic IoT type of business cases, but this is a very solid business case-- has been done for the last couple of years, very successful in a great number of industries. Basic idea is that on a vehicle, we embed a telematics unit. This telematics unit basically captures device data from different components on the vehicle. So for example, information about the state of the liquidity in the hydraulic components, measuring the noises made by the moving parts, et cetera. So this information can basically be aggregated by the telematics unit. It is then sent back to back end, to the vehicular operator, who can basically collect this data, analyze this data, and then start making decisions based on the results of this analysis.
And as we saw earlier, this might actually involve sending the service technician to the site where the vehicle is used, in a preventive maintenance type of case, where you actually send the service technician before the component actually breaks down, to basically avoid machine breakdown and insure that our machine equipment can be used 365 days a year, without any interruptions.
Of course, one key part of this is efficient management of data. So basically, collecting data from the assets, aggregating the data on the back end. And this immediately brings us to the big data discussion. Now this slide, I think is not going to come as a surprise. Everybody's predicting that in these IoT type of scenarios, the amount of data that will be aggregated is going to increase by-- is going to increase many fold. What might be a little bit more surprising, at least for us in our learning, was this really is a revolution. And with every good revolution, you don't know the outcome of the revolution, right? Ask Trotsky. What we saw is that the other thing, in addition to big data, really is flexibility and agility. And actually, in a lot of projects that we started, this was the first key driver. So the first concern really was on the project implementation, and not so much on the scale outside of things. So we just have to trust the good folks of Mongo, that the scale out eventually will work, and actually, we're glad to see that it did.
But in the initial project phases, agility and flexibility turned out to be really critical because we had to add devices as we went. We had to basically make sure that we could deal with multiple versions of the same devices and assets in the field concurrently. And that was the initial really big challenge for us. So about it 2011, when we started working on this, we saw that really what is required is a change in mindset with regards to data management technology.
So we looked at this-- this is actually inside. We've been working with a great company, Machina Research, doing analysis in the IoT fields. And I think this really nicely sums it up. And I was would have been glad if we had known all of this three or four years ago. But it turns out, in IoT, we're going from connecting 5,000 vehicles or 50,000 vehicles-- we're going to connecting millions of vehicles. It also means that we're not looking at the single purpose application, but rather really of a network of applications with various use cases, et cetera, that need to be loosely coupled and still integrated. Also, it means that we're dealing with all forms of data and all types of data processing-- so data streaming, real time analysis, batch processing, et cetera. So basically, the four key things that we were looking at were scalability, flexibility, analytics, and a unified view to these different types of data.
From our point of view, of course, yes, this is a lot about the data management technology. But this question actually goes far beyond this. And because we didn't want to reinvent the wheel in every projects where we needed to integrate assets and devices, et cetera, we developed a generic middleware platform-- this is our Bosch platform for the IoT. And in this platform, we differentiate between the three layers. So the first layer really is the device layer. So examples for devices are sensors, controllers, electric motors, et cetera. And typically, multiple such devices are aggregated into one asset.
So an asset could be, for example, a machine consisting of multiple components. And then, actually multiple assets can be grouped together into what we call systems of systems. So in this example, an assembly line could be such a system of system. And you need to be able to efficiently manage the data exchange between these different layers. So that starts with a means of extracting the data, finding a common format between different center types, between different data types, et cetera. Then being able to process this data actually close to the device. For example, for filtering purposes. Because of course, if you talk to the AT&Ts and Vodafones of the world, they will tell you, big data, we love it. Of course. Because that means you have to buy their data plans and data plans to actually collect all of this data.
But what you will typically have to do is you have to make a smart decision already on the asset, how much of this data you actually really wants to send over to the back end. So we typically try to differentiate between black, white, and gray. So sometimes there's a hard business case for actually capturing this data. Sometimes there's also a hard business case saying, no, sorry we don't need 125 meter reading a minute, or something like this. It's enough if we have aggregated this in chunks of 15 minutes or something like this. And then grey is something in the middle where we say, well, let's capture the data because it might be a business case, we're not sure yet, but cost is OK for doing this. And then of course we're aggregating all of this data in the back end. We're using MongoDB as our database management platform here in the back end.
And on top of Mongo, basically we have established a mechanism that allows us to aggregate this data. What's great, of course, is that yes, we use our data abstraction formats, but this is changing very frequently. And we can map all of these changes immediately to the Mongo database without actually having to go through redeployment cycles, et cetera. That was one of the first key learning for us-- very important. And then we built management functionality on top of this that basically enables us to monitor the status of all devices, also get additional meta data for all of the attributes of the devices-- for example, the time of reading. Because in many cases, the devices are not connected to 24/7. If a car has been in a garage or something like this, we might not have an actual reading for a couple of hours. So it's important at the back end to know how current actually is the data that we're looking at. So this is all functionality built into the suite. And we believe that combination of these two approaches is really incredibly useful for building IoT type of applications.
So I want to give you two examples. The first example is actually based on power tools. So I said earlier on I had a hard time believing that there is actually a business case for this. But it turns out, there are power tools that-- well let's say you and I would use. Bosch has the green line and the blue line. There are also power tools that not everybody in this room will have used in the past. So I'm talking, for example, about the Bosch Nexo tightening tool which comes at roughly $10k a pop. They are super advanced. They have a 35 volt battery. They support a power range between 3.5 and 15 Newton meters. That in itself is a little revolution. Because up until a couple of years ago, you were not able to have such powerful things.
And with this power now also, of course, comes the need to connect these things. So actually the next tool I'm talking about is a Wi-Fi enabled power tool which is great, because that allows us to basically hook this power tool into the manufacturer and maintenance processes as required. So tightening in itself is actually a quite complex process. You can imagine if you are looking, for example, at the screws and nuts and bolts used in an airplane-- you have six million of them-- and you need to control the process for each tightening process very accurately, because this is of course, mission critical. You also want to create a record of what has happened, so that if something goes wrong and insurance company's knocking at your door, you can basically say, well here's a print out of the six million tightening in our processes, right? So, go figure.
There's a lot more behind this. This is actually also changing the way how manufacturers are working. So we need to basically look at new ways for managing these mobile power tools. This is very different from a stationary tool embedded into a work cell. Well the first thing is, the stationery tool can't move away and get lost. So we need basic track and trace functionality. We need functionality that allows us, for example, to monitor the battery level of these tools. We need to be able to get remote readings on the calibration requirements of each tool in the field, and so forth. So this is what we're addressing with this first use case.
Technically, basically, we're using our proprietary device information model performance and actually, I didn't put source code on this, because it's going to be a long ugly hairy piece of XML. But eventually, this is mapped to a MongoDB JSON format. So in this example you can see, for example, the system information about the nut runner, it has an asset ID, it has a status, a battery status, and then a series of tightening protocols, for example. So basically, what we're doing here is we're use the Bosch software platform really as an enterprise service bus, so to say, really in the traditional sense. To integrate multiple heterogeneous devices into one single application platform, and then leverage what MongoDB as the central data hub for basically storing and analyzing the data that we're getting from these devices.
Of course, this has a lot to do with-- I said this earlier on-- with quality management, with time series capturing of the tightening processes, et cetera. So in the future, we're planning to do this in an even more advanced way. So we're currently looking at high accuracy indoor positioning technology that we can use in a factory floor, which basically will give us 3D information about the actual position of the different product components. So think aircraft, wings, et cetera, right? Plus, the actual 3D position of the tightening tool in the work floor.
And that basically enables us to do the following-- in the first step, the work program is uploaded onto the tightening tool. So the worker basically sees what he has to do next. So he walks over, and he sees this is the next tightening process I have to start. In the background, the indoor positioning system basically figures, OK, so now this is getting very close to the position of this particular product part. So now if the tightening process started, we can actually do a match between the tightening tool and the product, and create what we call a quality lot, that basically is a recording about this particular tightening step, that we can then basically capture right back into the MES or PLM system to make sure that we have 100% control over what actually happened as part of this process.
Another example I wanted to talk about is-- actually this is a literal translation of a German project name, that's why the abbreviation doesn't match the project name. So, Systematic Capturing of Fields Data-- this is a technology that we're actually building into a lot of the car components that we're selling. So, for example, the Bosch iBooster product, which is a brake booster, or power steering, the actual window wipers, all kinds of car components are enabled to support this technology. And we can then capture usage data from these different car components, really for two purposes. One purpose is actual maintenance of the individual car, to improve diagnostics. The other great value is, this is actually really telling us how the car components are actually performing in the field.
So were the assumptions that our engineers did about the specific design, for example, about the looser right? Are they validated in the field? How can we improve our products? So this is incredible value both for the OMs and for us, as a component manufacturer. And for us this is a small first step and in this revolution. So we've rolled this out as a pilot project. And actually, I come from a small company background. I'm relatively young. I've been-- sorry, I'm relatively young with the company, I meant to say. So I've been with Bosch now for 3 and a half years. And sometimes, these things just blow me away. So this is a pilot from a Bosch's point of view that has been rolled out in a small scale. Which means we have four pilot projects going on. The first one has 3 million cars, the second one 400,000, the third one 200,000. And this is what Bush considers to be a pilot project, right? And so far, everything's been working out great in terms of the whole process, also the use of Mongo as a core technology. As I said earlier on, for us it was really vital to make sure that they can go step by step, add component by component. And then also now, in the pilot, later and real roll out, to actually scale this up.
So let's just briefly look at the technology. The actual data capturing is done in different ways, depending on the customer we're working with. So some of you might know Bosch is operating a global network of car repair shops. We have about 14,000 of these car repair shops, and we are now gradually basically enabling them to actually also serve as data capturing units. There's also another approach for specific class of car, where the owners are willing to basically connect the car to Wi-Fi. So every time the car reaches the home base then the Wi-Fi can be used to read this data from the car and transport it back in real time. In both cases basically, we can use our asset management as an abstraction layer to capture data.
Then we have of course, the data stream processing. And I'm sure we will learn this week a lot on MongoDB's take on stream processing and where all of this is going to go. So we think this is actually a really, really vital piece for IoT type of applications, and we're very excited about the developments that we see here at Mongo at the moment in this area. Then of course we have collection, big data management, and analytics. In analytics, we really look at it from two angles-- there's the more let's say traditional, math lab type of analytic take on this, and then of course, there's the directly database embedded type of analytics that we're looking at.
So for us, so far this has been working on great. We're building this out as a software, as a service, working with multiple customers, as I said early on. And as I mentioned, the key thing for us was we need to use a combination of structured, unstructured, semi-structured and polymorphic data. And whoever is able to give me a very precise definition of these four things, and tell me that they're actually not maybe like two or two and half things, talk to me afterwards, OK?
Some examples here-- unfortunately, I had to black out some of the real customer data, OK? Because this is very confidential. But here you can see this is adjacent structure where we can basically see different parameters that we are basically storing in this project, OK? So unfortunately, it's also it's German, but it will give you some examples. Well, some people like German cars, so I think it's OK. So we have the AIF dates, we have the powertrain type, we have to reading date, [INAUDIBLE] basically means, when was this particular series of cars built. We have [GERMAN], which pretty much means, it's the ID of the actual capturing of this data set.
We can see information like what's the kilometer range of this particular car. On the right side, there's one particularly interesting data entity, which is VIN, which is the vehicle information number. And this is actually where it starts really getting interesting, because this relates to the whole question of who actually owns the data, right? And for us, as Bosch, this is maybe the most important question. So that's why in this particular example, we are actually not storing the car ID, but we are using a hash to basically anonymize this information, but still be able to basically, if the same car comes back, match the readings with the previous readings.
So we believe that this is really important that at the end the day, the customer can make the decision about who should own his data. This is another example, for multidimensional data gram readings. So here you can see we decided the data modeling to basically have different ranges. So there's for example, on the right side, there's a range from 1 to 6 where we have different pressure readings and different groups to manage this multidimensional data capturing. And what we see is that the current approach really is great for us to enable different types of diagrams, different data ranges, to capture geo-information, and so on, and so on. So really important for us that we can handle this on all the required levels.
And that already brings me to the end. I'm sorry. I didn't keep track of the time, but I hope that's OK. So the message I was trying to bring across is IoT-- well, I wouldn't be here if it wouldn't be exciting for me. But it is happening, it is happening step by step-- but with a huge momentum. And in order to really handle this step by step adoption, we need flexible, and scalable technologies. And we believe the combination of Bosch technology and MongoDB technology is giving this to you. So if you actually happened to have any needs for IoT technology, beyond soccer goal line control or something like this, come to me afterwards and talk to me. Thank you.
Thank you so much, Dirk, for sharing that vision with us about the revolution. It certainly seems like, even though we don't know the outcome, that you're doing an awful lot to be making us safer. I do have to say, however-- power tools and not a connected refrigerator? Not so fair. Does anyone have any questions for Dirk, please? [INAUDIBLE].
Unfortunately, no. That's part of what I said earlier on, customer confidentiality. And this is what Bush stands for. So, sorry.
Hi, there. What's interesting is-- I'm right here.
Sorry.
What's interesting is, every story you tell about your use case, everyone can immediately relate. But what I'm interested to understand more about is how did Bosch as a firm decide, we know there's all this data out there, how do we as a firm really become data ready and view data as an asset. So what happened internally at the firm where the company says, OK, this is a really, really big thing. Let's get our arms around it, and get to the point where we can create air bags with sensors and power tools with Wi-Fi.
Yes, of course, being a German company, we did a very structured approach. We made a plan and then we did it, right? This all started about five, six years ago, when our former chairman, Dr. [? Deis, ?] at the time basically said, OK, this could be the Kodak moment of our industry. And he carefully put measures in place to start addressing these things.
So, one of the measures was actually to set up a legal entity, that would drive all of this within the larger Bosch organization, to equip it with funds for let's say, careful M&A activity. So we acquired two companies that add great value to our IoT platform. And then also of course, start and internal organizational transformation. OK, so Bosch software innovations is not the only part of this internal transformation, but of course, we try to help a lot with the different Bosch business units. But also external customers-- I think this was also another key piece, that from the very beginning, we said, OK, we can not only focus on our internal transformation, but we equally have to look at what's happening with our customers, and what's happening with us. And the internal adoption of techniques like open innovation, the setup of cross-business units, work groups, to make this happen. And so I think it's been very good so far. We might not-- and you can't quote me on this-- we might not be a Google that fork out $3 billion to do something. We do this very specifically in the Bosch way, and what I just described I think for us has been very good so far.
And how would you describe your relationship between the engineers and the technology people and the product mangers. So the product manager in charge of the drill talking to the engineer that says-- Who has the vision? Is it coming more from engineering, is it coming more from products? Is it a collaboration between the two?
I would definitely say it's a collaboration between-- well not only the two. It's really a collaboration from the very top of the company, so our group CEO, Dr. Denner. Really is standing behind this and supporting this, down to the level of, business unit managers, product managers, and even engineers. Just last week, I got an email from an engineer saying, listen, I saw your presentation on IoT the other day. I have some spare time on my hands, can I do something? For me, this is really amazing.
Hi, I'm Joseph Cocoa with Vanderbilt University. You had mentioned that you were curious-- or I'm sorry, you were deciding whether you want to aggregate data from individual objects. At what level, what granularity, essentially from the business standpoint, you want to record that? I was curious if the devices are actually aware of how often or what granularity their data is going to be recorded at, or how that information gets proliferated-- where that process begins essentially.
I think this is where this perspective basically comes in. We have to deal with very simple devices. So the soccer ball I was talking about early on-- super simple, important to many people, but still-- like really like a tiny thing built in. So around this, we actually have to build other things, like magnetic field reading component, et cetera. And all of this might actually form an asset, which is the football field, right? And really always in our experience comes down to, hierarchy of very simple devices, more powerful assets that aggregate multiple devices, and then up to what we call, systems of systems. And this is when you have to manage, in our experience, this layering.
At what level on that layering does it issue a command of saying, I want to aggregate the data and not look at it. [INAUDIBLE].
On pretty much every level. So for example, we use a rule technology that we acquired on these different levels to actually look at the data and make decisions on every level. Because some decisions, you can immediately make down on the device level. For other information, you need context information that is only available in the back end. So you might actually say on the device level, OK, there's a 50% chance that this is interesting to the back end, so let's send it over. And then on the back end, I look at this in context and say, nah-- not at the moment.
Last question.
Hi. I actually have an important question-- in '66, had goal line technology been there, would England have beaten Germany in the World Cup-- No, I'm just I'm just kidding, of course. So are you using Mongo-- or what kind of analytics that the essence of, the nature of the analytics that you're doing in the back end right now-- is Mongo the complete solution, or what other tools are you looking at? Like, GE, for example, is building their industrial internet, which is similar in scope, and I guess ambition to what it is that you all are doing. And there's an awful lot of effort being put into looking at machine patterns, particularly around being able to predict reliability and availability. Is that one of the scope features of what you're working on? And what kind of tools are you employing for that type of work?
I don't think there's a one size fits all-- answer to your second question. So, on the database level, so we're looking at multiple technologies. We are looking at, for example, at Hadoop, for integrating multiple resources, et cetera. But it's not only a question of database technology. There's for example, one partner company that we work with, National Instruments, who have great technology for what they call analog big data, which is actually a little bit further out, before we actually go to the database level. And you need to combine all of these different approaches very specifically to your particular problem. So there's no one size fits all answer.
OK, thank you for that. We appreciate your time. Everyone, please make sure you fill out your survey at the end of this for Dirk and his great presentation. In about seven minutes, we'll start again with the AHL Group. They're going to talk about their transformation from traditional systems, and all of their financial data into MongoDB. Thanks so much.