Harel Kodesh, Vice President, Chief Technology Officer at GE Software gave one of the most impactful, no-nonsense, keynotes at the Cloud Foundry Summit.
Kodesh introduced a fresh perspective on Internet of Things (IoT), contrasting the familiar Consumer IoT with Industrial IoT, or what he calls the Internet of (Really Important) Things. His talk paints the picture of how important the Industrial IoT is to large manufacturing and supply-chain systems, with far reaching impacts across governments, cities infrastructure, aviation, power generation, transportation and more.
|Commercial Internet||Industrial Internet|
|Data||80GB of data a day at Twitter||500GB of data for one airplane flight|
|Security||Non-critical upsell to consumers||Compliance mandated by government: HIPAA, ITAR|
|Data Privacy||Lack of privacy somewhat acceptable||Lack of privacy not an option|
|Connectivity||Connection interrupts might lead to a dropped phone call||Connection interrupts might lead to an oil spill|
|Device Lifespan||6 months – 2 years||25+ years, capital assets|
Blazing the path forward to the industrial internet, GE scaled to nearly 20 Billion sensors in 2014. They expect to be producing an estimated 80 exabytes by 2019— all of which needs to be ingested, stored, cleaned, and analyzed in order to gain insights about how to efficiently operate and maintain their capital assets.
Building on Cloud Foundry as a scalable, reliable and available platform, GE Software builds apps that deliver analytics, business insight, and solutions to their customers. In addition, GE extends the Cloud Foundry ecosystem by building and enhancing protocols that enable a diverse set of IoT devices. These protocol innovations are planned to be contributed back to the Cloud Foundry community.
- Missed Cloud Foundry Summit? Read the recap here
- Watch more Pivotal sessions
- See all sessions from the Cloud Foundry Summit 2015
- Learn more about Pivotal’s perspective on IoT
Good evening, good morning. Let’s see if I can get this thing to work. That will be my first test. Some of you are wondering what a GE executive is doing in a conference like this. It’s almost like a crop duster coming to an organic grower conference. Let me take the next fifteen minutes and try to explain what we’re building and maybe by the end of the talk I’ll be able to convince you that we are working on the same thing that you are working on.
Internet of things, IOT, is a very important thing. You can see it everywhere, everyone is talking about the numbers of devices that are showing up the story that is going on. You have four or five very important companies working on that. Google, Amazon, Apple, maybe Microsoft sometimes, when they feel like it.
Overall, these are very important companies and they have about nineteen percent, not only that, they have only eight percent, but overall the software industry is all about nineteen percent of the S&P 500 total market value. Very important.
Egg cookers that connect to each other, communicating tooth brushes, belt buckles that light up, all kinds of interesting things. This is not what we’re talking about. We’re talking about something much more important. You may not feel it, but at the end of the day when you open the faucet, you’re using GE equipment at the other end of the line. When you turn on the lights, there’s a turbine at the other end that actually produces this power, this wattage that you’re using. If you’re taking a flight, which some of you are going to do after I’m done and the next speaker is going to be done, you’ll be flying, seventy percent chance, a GE aviation product.
So overall, we’re talking about big, dirty, noisy, machines that are very important for our life in general, our well being specifically. Now, the interesting thing about that, is all those things have sensors and they talk day and night. Twenty-four seven. Somebody needs to analyze them and make sure that we understand exactly what’s going on with them.
If you look at the numbers of the sensors in those machines, we’re approaching the point where there are twenty billion of them. They create about eighty exabytes of data. Now, this is a lot of information and the reason we want to collect it, is not to break any privacy rules, or do anything of that sort, but really to understand how we can optimize those asset potentials to generate more revenue. How to make sure the production and the livelihood of those countries that are actually working with those assets, much smoother. How to make sure we can extend those assets and allow them to generate revenues in the years to come.
Those things have actually been going on for a very long time. Some of those assets were put in place thirty or thirty-five years ago. Naturally, they didn’t have all the controllers and the computer environment that we’re looking at today. The goals of GE and the companies that are working within the same ecosystem, is to make sure that if we really build this thing we called industrial internet.
You have the consumer internet, you have the enterprise internet, Salesforce, Splank, and so one, and you have the industrial internet. The industrial internet, that is actually going after that part of the GDP, that is little bit of an underserved community. It’s not something that every day you wake up thinking about how a turbine talks to another piece of equipment on that internet.
We would assert this is really tomorrow’s growth opportunity. There’s a big difference between the two. At one level, they don’t look that different, but actually if you think about it, there’s a huge difference in several key dimensions. That’s why we’re here today.
First, the average longevity of those devices, and no offense to the tooth brushes and the rice cookers, is that usually people use them for about six months. It can be a nice [inaudible 00:04:02] that you bought, it can be a nice gadget, these things are working for an average of six months. Our assets need to work day in and day out for about thirty, thirty-five, years. When you think about that, it actually … A lot of the operators have to make those decisions about whether they take the drill into maintenance, or forget about maintenance and keep drilling oil because the price of oil is jumping by five dollars.
Those decisions are really deep, economical decisions that people have to make as they go and work those assets. The other part, connectivity. The biggest problem that we can have is the call will be dropped, there’s definitely a certain loss of productivity when people can’t talk to each other, but when you have a blow up preventer that you have to overhaul or upgrade in the middle of drilling operations, and you know that the blow up preventer is not going to work the way it should. Then, God forbid, another oil spill is going to be upon us.
This is where connectivity is becoming very important. Not only do we have to make sure that we will be able to get this connectivity all the time, we also have to make provisions for when the connectivity is intermittent, or really to some extent, unavailable. If you’re in the middle of the sea, it is usually unlikely that you’re going to have a cell operating in the vicinity. You have to rely on satellite communications, you have to optimize the cost of that, you have to make sure that under any circumstances, the industrial assets will be able to keep working.
Big difference. Data management. Let’s talk about how many exabytes we need. If you think about that, there are about ten million … Something along the lines of every second of tweets that are happening on the network. There’s about eighty gigabytes a day. This is a very respectable number. If you look at Facebook, you’re talking about hundreds of petabytes a year.
Now let’s look at what happens when you deal with GE aviation, and this is on GE aviation. Five hundred gigabytes for every flight. It’s all data about the engine, the vibration, the cabin temperature. It’s important for mechanical engineers to build the air frame, it’s important for GE aviation folks that build the engines, and it’s import for the marketing and the performance scientists in the airlines who would like to understand whether the cabin was too warm or too cold.
Now about one hundred thousand flights every day happen in the United States. If we only took the part that we are dealing with, that GE is dealing with, which is about seventy percent market share. We’re talking about seventy percent of one hundred thousand flights, times half a terabyte. I think in my math it’s about fifty petabytes a day. We’re talking about an unprecedented amount of information and for the first time in history, people actually know what to do with all this data.
If you go and analyze that, you’re going to figure out how to build better engines, faster engines, how to make sure that the blades, the turbine blades, can actually last much longer. You have an end to end picture of everything that we’re building.
Security. Everybody wants the devices to be secure but would you pay one hundred dollars more for secure phones. Maybe if you’re in the intelligence community or in a business that is exposed to industrial espionage, that might be the case. Normal consumers don’t pay for that.
Our industry will pay a lot to make sure that their devices don’t get hacked. I don’t think I’m telling you any secret, but you can actually hack into pretty much everything unless there is enough security in there. If somebody goes and fucks around with a turbine or an MRI machine or a CT scanner, this can actually have some very important and very unfortunate consequences.
Most of the things we’re building are considered national, critical infrastructure. They have to work all the time. We want to make sure that at the end of the day, the industrial internet is not only high performance, but also high security. We cannot tolerate breaches, either by nation states or bad people that would like to play around with some of those assets we have at our disposal.
Zuckerberg said several years ago that privacy is no longer the social norm. To some extent, while some people were upset about it, it is true. We are much less private today than we were twenty years ago. It’s okay to put things on Facebook, it’s okay to tweet whatever the heck you think about. At the end of the day, when it comes down to a different set of devices, again the industrial devices, the medical equipment, this is where privacy is not just a matter of sentiment, of somebody or a doctor or a patient, it’s a regulated and required property of the over all system.
You have to be HIPAA compliant, you have to be [inaudible 00:09:15] compliant, you have to have sold twenty-seven thousand. You have all these standards, some of us may or may not agree with the fact that they actually have some sort of burden on our system, but it is what it is. The politicians, the people, the governments, have all decided that this is how this set of equipment that we’re building has to behave.
We have to make sure the industrial internet, while at the risk of being a little bit more bureaucratic, is something that enables us to really serve and comply with those standards. With that in mind, we have to build something that will allow us to take all this data, make sure it’s secure enough, it’s compliant with the right standards, ingest it, store it, clean it, and make sure that at a certain point we can take all of that, analyze that and come back to the operator of those assets with insights.
What is it they need to do that will make the operation of those things better, more profitable, and prevent the bane of the industry that we’re dealing with, that is called unplanned down time. Unplanned down time is when you sit on an airplane and the captain comes in and says well this plane is not going to fly today.
Every body gets out, everybody goes and finds a different plane. This is something that was not planned. If it were planned, then you wouldn’t’ be sitting in this plane, it should be sitting in a hanger getting worked on. At the end of the day this is what we’re trying to optimize. We’re trying to optimize the operation of assets, we’re trying to give our operators some insights about what is it they need to do.
So we had to focus on the higher level of the stack. For that, we had to find the right lower levels to make sure that we can do it at scale. GE is working in one hundred seventy countries around the world. We have to make sure it’s reliable because our customers don’t expect anything less than twenty-four seven. Four nine, sometimes five nine and we want to make sure that we can get something that can allow us to push the envelope, as we build the industrial internet for years to come.
This is where we started working with Cloud Foundry. It was not a natural journey, because for GE, a company that has been doing what it’s doing for one hundred thirty years, to go and change everything we know about software, get into Cloud Foundry culture. Working with people in other places in the ecosystem, was not very simple. What it did, it allowed us to move a lot of the people that were working on that level in the past, and deploying them on a higher level of the stack so we can work on the analytics.
We can work on the way to ingest and keep the data, and make sure we can get into the right data at record time, without sacrificing the reliability or the completeness of that. Some of the data that we have to keep, we keep because we like it and because we think it’s going to be good for us to understand what the assets are doing. Some of the data we have to keep because we were told to keep it. The government told us to keep it. This is regulated data that we cannot get rid of.
So Cloud Foundry was paramount in our ability to lift the discussion and lift the focus of what we’re doing, to a completely different level, that will get us closer to the industrial internet abstraction. What is an asset? What is a locomotive really? When you think about it a locomotive is a big thing running on railways, but at the same time it’s one hundred and fifty thousand parts, traveling together. Sometimes you have to look at the locomotive as the big hunk of iron, and sometimes you have to look at a part and have to figure out, where is the right depot, where are the spare parts, and where are the other locomotives that this part is installed in?
All those things are things that we have to think about and solve ourselves because there’s nobody in the ecosystem that will do it. Other things, we rely on Cloud Foundry. A case in point is a wind turbine. If you look at a wind turbine, it’s a magnificent thing by the way, if you get all the way up it’s awesome, but if you look at that, think about what we’re doing, wind turbines, it’s like a wolf. It never hunts alone, it’s never deployed alone, usually you have a wind farm. An interesting thing about a wind farm, when you have all this wind that is coming to you, you can optimize the operation of a specific turbine, but by doing that you’re going to create turbulence that will compromise the operation of the turbine behind that.
What you want to do is to create analytics that will allow you to optimize the performance of the whole farm. Some of the execution will have to be in the system, some of the execution of the analytics system will have to be at the controller level, and some of the execution has to be at the specific turbine. All those things have to be dynamically balanced because sometimes the issue is local, sometimes the issue is regional, sometimes the issue is … Wind and everything else is much more sectorial in that respect.
The data that you get … You collect data, if you look at it, at time series data, it’s pretty much every fraction of a micro second. There’s a lot of data to ingest, we have to put it somewhere, we have to analyze that. That’s what Cloud Foundry does for us, that’s what Cloud Foundry enables us to do as we go and solve the predicts problems.
To that extent we had to do two things. We’re excited to do two things. One is to build the incubator for the industrial protocols. We have a lot of data that we need to generate. Those pieces of data cannot just flow on an HTP channel, which is what we got from Cloud Foundry, but in the best spirit of open community, we are working on the ftp of protocol routing, in such a way that we’ll be able to take those industrial protocol and move them into the cloud.
We are contributing the results of that project back to the open source community. The other thing, looks like every person needs a dojo, so we have a dojo too, and this dojo is all about the industrial internet. Industrial internet is too big for GE, even as an one hundred and fifty thousand people company, to take on itself. This is something we’re opening up on the other side of the bay, and the local headquarters in San Ramon.
What we’d like to see, and what we see now, other than people in GE and some other distinguished members, we would like to see representatives of the whole community coming to work with us. Really pushing the envelope on what we call the industrial internet.
At the end of the day this is just the beginning. Now that I’ve explained why GE is here, I’m looking forward to being here again next year, or any other time in between and actually working with you folks, and anybody else that wants to come and join us on really pushing forward this industrial internet. I think it’s going to be great. Thank you very much.
About the Author
Pieter Humphrey is a Product Marketing Manager responsible for Java Developer Marketing at Pivotal Software, Inc. Pieter comes from BEA/Oracle with long history of developer tools, Java EE, SOA, EAI, application server and other Java middleware as both a marketing guy and sales engineer since 1998. Find me on Twitter at https://www.twitter.com/pieterhumphrey.More Content by Pieter Humphrey