IoT and Resilient Systems

May 3rd, 2016

We’ve recently been getting more and more questions about the “Internet of Things” (IoT), both from conference attendees and clients. What will the killer app be? Where will the value be found in a world of connected devices? How should my company be thinking about this? Many conversations begin focused on the things themselves. What’s your favorite thing? Are connected coffee makers going anywhere?

As someone who has come home to a smoldering domicile and an accidentally-left-on coffee maker, I can appreciate the value of being able to remotely turn off your connected appliances. But, it turns out, that was not what caused my house fire (correlation does not equal causation), and that may not be the killer app for the IoT either. We believe there are clearly some compelling value propositions that come from integrating the visibility from the IoT into applications that help understand and manage the state of complex systems. With the internet of things, the more things, really, the merrier.

The Value of IoT

You can split IoT value propositions into two varieties: the value proposition the connected device itself provides to its user by virtue of function and connectivity, and value propositions that take advantage of a network of connected devices to serve a community of users. A very simple example of this is the value of the maps app on your iPhone, compared to the value of Waze. The maps app can help you find a route to a destination — valuable to the user. Waze can help you understand how long that route is likely to take based on the data from the entire community (millions) of users.

Recent reports demonstrate the profound potential value of connected devices to their users. A Fitbit saving the life of its user clearly has tremendous value to that user. On the other hand, the network of Jawbone users provided a very interesting view of the impact of an earthquake. The value of the network’s data in guiding emergency responders in an earthquake is in aggregate also quite profound.

Beyond the aggregate value propositions of a network of IoT devices, there’s an additional benefit of having a rich network of connected, sensing devices at our disposal. Such a network allows us to build resilient systems which gracefully degrade when signals are unavailable, rather than outright failure when a single source of visibility goes dark. Think about your car: when you want to know how fast you are going, you sensibly look at the speedometer. But what if that fails? Well, you could probably use the combination of the tachometer (how fast the engine is spinning), and position of the gear selector, to get a reasonable approximation of how fast you are going.

We have used this approach to build gracefully degrading systems that help us understand how the Caltrain is running. There are lots of sources we use to understand where a particular train is at any given time — a key element of our application. Any individual source of GPS data about a train, be it the data coming from the locomotive’s GPS capability, or from a user’s smartphone, is unreliable and error prone. By combining signals from users actually riding the train, Caltrain’s API for expected arrival times, Twitter, and our own direct observation using audio and video (to name a few), we can reason about the state of the Caltrain system even when individual sources of visibility are experiencing outages. We can also better validate signals by comparing across sources.

We believe that these network effects of combining signals from a rich network of connected devices — the internet of things — present opportunities for tremendous value, and resiliency. For a deeper look at our Caltrain work and the internet of things, check out this talk from the Data Science Pop-up Conference in Austin. To learn more about how we use these signals to understand the system, read this post from Tom Fawcett and Christian Perez.