First of three parts: Understanding the usefulness of data, what to keep and for how long; focusing on solutions rather than just chips; networking issues caused by a flood of data; unexpected uses for data and technology; economic considerations.
Semiconductor Engineering sat down to discuss the future of the IoT with Oleg Logvinov, director of market development for STMicroelectronics’ Industrial and Power Conversion Division; Martin Lund, senior vice president of the IP Group at Cadence; Naveed Sherwani, president and CEO of Open-Silicon; and Damon Hernandez, a member of the Web3D Consortium. What follows are excerpts of that conversation.
SE: What are we going to do with all the data from the Internet of Things?
Lund: First of all, what is it? There’s the notion that it’s an Internet of things rather than people. In the past, the Internet has been about connecting people with information that can be consumed. It may be generated by things, but it gets consumed by humans. Big data is an offspring of that. We’re creating all these data centers around the world to process that data. Seventy percent of the traffic on the Internet is video. About 1% to 2% of all energy is consumed by data centers. So if we now can increase that data 1,000 times, because we’re going from 1 billion people to tens of billions of things being connected and communicating all the time, it’s a huge increase. What do you do with the data? First of all, you don’t have to remember all of it. If a sensor tells me it’s 70 degrees in my home and it was the same two seconds ago and four seconds ago, it doesn’t mean I have to save it on my hard drive. I can ignore all data that is not useful for me. That helps, but it probably requires a paradigm shift in the way we think about data because a lot of people don’t like to throw it away. You actually have throw it away in this world.
Logvinov: But in some cases, what you throw away today might be useful in the future. It could be a gold mine. You don’t know what data is valuable today or tomorrow. When you really think about the pervasiveness of the Internet of Things, it’s the ability to communicate with things on a molecular level. It’s being able to tap into each neuron of our body and communicate this information. The amount of information we’re talking about goes way beyond exponential—forget about storage for a moment—just for transmitting and processing it. We have to think about micro-clusters that consume and produce data in some form, and only some replica of the data should be exposed to the cloud.
Sherwani: People keep trends, which is more useful. A temperature trend cuts down on the data quite a bit. That’s probably the most useful aspect of that kind of sensor. We don’t have to keep raw data.
Hernandez: Dealing with the standards for Web3D content—and everyone is looking at that as the platform for the delivery of this data—how do we make it manageable and how do we understand it? How is the user going to be able to take advantage of this data in a relevant way?
SE: We’re talking about adding a level of intelligence into the Internet of Things rather than just dumb sensors. But if you send all this data through the Internet, it’s going to bring down the whole system. How do we deal with that?
Sherwani: This vision that everything will be communicating all the time doesn’t work. Devices have to know when to communicate. We are working on a luggage tag. The question is whether that luggage tag should be transmitting all the time, or whether it should transmit when it may be lost. If it were to transmit all the time, you couldn’t afford the power that goes into that. There has to be some intelligence about the power involved in it. Only the luggage that is lost should be transmitting.
Logvinov: It’s not just how much data you transmit all the time. It’s also the issue of privacy. If you think about a camera installed in someone’s apartment. Would they be comfortable if the camera was transmitting live images to everyone in the world? Probably not. But at the same time it probably would be useful to have a camera in a home for the elderly, which could detect a fall and notify you. That type of data abstraction and extraction of meaningful events is important. When we talk about the Internet of Things, systems that only transmit what they are designed to transmit—and which can take out meaningful pieces of that data—will be useful.
Sherwani: People from semiconductor companies have a difficult time with this because the people we are working with have no understanding of semiconductors and zero interest in them. They are looking for a solution, and on our side we are not as used to talking about complete solutions. But we do have to go in that direction, and those solutions have to be very low cost. If you have a system solution, then you have a huge margin to work with. It’s a very challenging business model—provide complete solutions at a low cost.
Lund: It’s a networking problem at the heart of it. What we’re staring at right now is where we were 25 years ago. People are thinking about the cost of transmitting useful bits. This notion of sending tons of data that is streaming out across the Internet in a flat network won’t happen. There isn’t enough bandwidth. And we don’t need to have a thermostat in China talk to a car radio in Norway.
Hernandez: It’s very dependent on the user experience. What is the purpose of that data being sent? It doesn’t always have to be on because you may only need that information every 3 minutes or every 30 minutes. The user experience is going to dictate how often we’re querying these sensors for their data. They won’t always be on.
SE: If you have a sensor in your body, an individual reading may make no sense but over time those readings may say something important. You don’t necessarily recognize the value right away. Don’t you still have the issue of what to do with all that data?
Sherwani: How did we deal with this earlier? Security cameras have been around forever, but they only keep two weeks worth of data. The security camera is needed when something happens.
Logvinov: What you’ve described is a point solution for a very specific purpose, and not open to communicate with anything else in the world. Two weeks is perfect in that case. But as we transition into the Internet of Things and we begin talking about intelligent things, those things are going to be used by multiple services and systems. If you think about your smart phone, it’s a very good example. It has a camera, it has a location device, accelerometers and a bunch of things. That goes to a large number of service providers to take that data and do something different with it. This is a key transition, from a point solution to a general-purpose device capable of gathering data. It’s very easy to define a use case for a point solution where nothing beyond two weeks of data is necessary. But for a general-purpose device, data for a medical service, imagery to a storage facility because maybe later on this image may indicate the progression of some disease—that could be relevant. We need to start thinking about where the data will go. It’s not an issue of storage. Storage will become cheaper and cheaper. What we are not keeping up with is how to preserve the privacy of this data. How do you create the layers that allow you to filter the privacy levels you want today and tomorrow? That is the element missing today.
Sherwani: On data storage, every smart phone has 70% more storage than the user will ever need. All the cell phones being shipped now are being shipped with 70% more storage than they will ever use. From a storage point of view, that’s not a problem.
Logvinov: I remember 20 years ago when someone said we will never need more than a 20-megabyte drive.
Lund: At some point we run out of electrons. It comes down to what is useful. At the end of the day, it’s economics. We’ll keep as much data that is useful, and that may change over time. But the Internet of Things is only useful if it generates economic or human value, which may be in the form of managing greenhouse effects or energy across a country. Otherwise it’s just interesting data.
Logvinov: Five years ago everyone thought the location of wireless access points was completely useless data. Look at what we’re doing today with this data. We’re helping our GPS find itself by locating nearby WiFi.