Manufacturers, utilities, health care and other industrial and services organizations have an opportunity to develop applications that model and draw upon the capabilities of the increasingly connected physical world around them. Seems, after all, as if almost everything already is or soon will be connected to a sensor of some sort, reeling in data to private intranets and, phase by phase, to the Internet, and creating opportunities to create smart grids, smart parking, and smart cities.

Thingworx may be able to help them take advantage of that opportunity. The start-up today plans to formally launch its application platform in Downingtown, Pa. (hopefully bringing a bit of cheer to state residents still getting over last night’s Pittsburgh Steelers Super Bowl loss). It leverages its semantic definitions for this “Internet of things” world to help those organizations – and not just the techies within them – to search, query, and analyze data, and then build mash-ups using the results.

“Our search and query tools use semantic definitions to help people discover their way through information,” says CEO and co-founder Russ Fadel, who also was CEO and co-founder of Lighthammer Software Development, later acquired by SAP AG. It covers how entities, such as “battery charge,” are classified and what property type they belong to, which fields are queryable, which fit in maps, which are keywords, and so on. Non-technical users then can use Thingworx to build mash-ups using the semantic model, drawing upon connections from their search results: Take all electric cars of type Toyota within 100 miles of a certain location that have a battery charge level less than 20 percent and are currently running, and plot them for display on a map, for instance.

As Fadel puts it, “we don’t tell people they are building a semantic model, but that‘s what they did, to interact with data and functions of systems in ways that were almost impossible before.”

How Semantic Is It?

So how semantic are Thingworx’ internal concepts of definitions and its approach to helping customers discover and integrate data from various systems, for creating applications dependent on leveraging the richly described services, data and events that surround things?

“The world we live in is farthest out on the horizon of something you can imagine having a standard set of semantic definitions for,” says Fadel. For example, industry agreement about how to define and measure the performance and operation of what the exact same piece of physical equipment does, such as a beverage bottling machine, in two separate soda companies is hard to come by, thanks to stakeholders’ own specific views of bottling line processes.

A lot of what the Thingworx solution does, the company says, is build a meta-tagging and vocabulary system on top of a lot of different connected things – of the representation of the real thing itself, such as a car or truck; on the data it generates, such as exhaust or real-time status; and on human-generated content related to it, like blogs or Wikis. All of these can participate equally in the collaborative process –a machine, say, can initiate a discussion forum topic on a quality issue its sensors detected and notify appropriate parties. “Maybe that [message] is meta-tagged with the production order or the product they were making at the time,” says CTO and co-founder Rick Bullotta. Bullotta was a fellow co-founder and CTO of Lighthammer.

The complexity to give access to all the data it represents in a uniform way is tough to fit into the semantic web standards model, Thingworx says. We’re looking at some 1 trillion connected devices predicted by 2025 to add to the fun. But even before the Internet came to be as a mechanism to connect the world together, systems like those bottling machines in manufacturing plants were wired up to send along data on private or closed networks. These data sets went into proprietary data stores, so customers generally are not starting with a clean slate when they want to try to get some value out of that data.

“It’s challenging to project some of the existing standards to do the totality of what we need to do,” says Fadel. When the customers have existing information assets, Bullotta says, “just in general the application of the semantic web to legacy is a challenge.” Trying to automatically infer appropriate semantics and meta-tags is an approach that’s been tried, “but it’s a hard problem. So we had to have an approach that fits all the oddball legacy stores we encountered.”

Thingworx has a graph model database under the hood to accommodate hierarchical, tabular or time series streams of information – and some scenarios not envisioned yet, either – that has some built in RDF and lite SPARQL, Bullotta says. But when it comes to user-friendliness, Thingworx finds, for instance, that SPARQL is great for programmers but still too awkward and difficult a way to express content services for normal people, he says.

“In general SPARQL is the point of sword for making the decision to leverage some standards, and it couldn’t fit how we wanted to express queries,” Bullotta says. “So we are still storing stuff in triples but we have chosen to focus on a much more accessible way to get at that data for a non-technical person because our focus is on end users. But because that layer exists, if we do want to expose that information in the RDF model, it’s not a stretch to turn on that functionality.”

But there’s plenty of hay to be made in the meantime. “We have huge value to bring in the Internet of things,” he says. “But we are pragmatic and know there’s a huge amount of value to unlock today in the existing networks of things.”

Photo credit: Flickr/ J. L. Trinh