Two startups break through the computing fog
Fog computing. It’s one of those phrases I detest because it’s such marketing jargon. But I don’t have anything better to offer, so I’m going to use it.
When people talk about fog computing, they are referring to this idea that there will be lots of sensors or computing devices spread about a factory, a home or wherever sending data to a gateway device for preliminary data processing of some sort before sending it to the cloud. Fog is a low-floating cloud, hence the name for this type of architecture for the internet of things.
With this architecture, which has been popularized by Cisco, Intel, Dell and others selling gateway boxes or services on those boxes, is an opportunity to develop a new suite of software to handle data. So we’re seeing a lot of startups try to enter the fog with new ideas about how to handle, analyze and stream data.
Two I’ve recently encountered are Flowthings, which has raised about $10 million from investors including IBM, and FogHorn Systems, which recently raised $12 million from investors including GE and Bosch Ventures. Both are building software to help handle data processing inside these gateway boxes.
Since fog computing is a developing architecture, both companies are hoping their technology becomes part of the future essential “stack” for fog computing, much like Linux became part of the open source stack of technologies underpinning the early web and Spark and Mesos are two of the core technologies that comprise the real-time data SMACK “stack.”
When it comes to fog, there is still a lot of debate over how exactly to handle the data. Some companies view a gateway as a place to start processing data and normalizing it. Some think of it as a place to perform some pre-processing that can then be enacted on the site. That sort of philosophy works well in places where data connectivity is limited or you need an immediate action.
But both CEOs of these companies are adamant that you need a different way of thinking about your data. You have to view your data as it’s coming in. David King, CEO of FogHorn derides the idea of a data lake, this massive storage of all the data a company collects, as being of much use. “You have to tackle that preemptively,” he says. And that means it’s tackled in a box at the site.
Flowthings CEO Eric Alterman agrees. “Once you look at data as streams, the idea of parking that in a database, even in-memory, it’s the wrong metaphor. Streams are data in motion and you keep it in motion until it does its jobs and comes to rest.”
This is all well and good, but the challenge then becomes how to process what can be a torrent of small data coming in from thousands or even hundreds of thousands of sensors. A software provider has to be flexible enough to handle that data while also not running the software in a large footprint. Many of these gateway boxes are full-sized computers, but they don’t have the infinite resources of the cloud.
Both FogHorn and Flowthings have built code that runs in small footprints, with Alterman hoping to one day be able to put some of his code on sensors if needed. And both have turned to reactive programming to handle the challenge of processing data in motion.
The idea is that instead of storing data in a database or near the processor in memory, commands are applied to the data as it runs through the gateway. To make what is incredibly complex under the hood simple, Flowthings has turned to a model where new jobs are “wired” together using a GUI interface. This lets developers build new rules on the box but also brings the same programming model to the cloud.
FogHorn uses expression language to tell the data what to do in real time. Both CEOs say that they realize that this concept of making sense of data in motion will result in an architectural shift, and both are excited because the industrial internet is a perfect use case for their goals.
Right now they both work with the systems integrators and gateway box makers to find customers. Alterman says he works with Dell because its Intel gateway “has a lot of pep.” Given their intermediary, but crucial role in the industrial internet, it’s likely we’ll see them snapped up by a systems integrator over time.
Did you like this story? Want more? Sign up for Stacey Knows Things, a newsletter covering the internet of things, to get this essay and more.
