TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

Q&A: Tremendous Benefits Ahead in the Internet of Things

Connecting physical objects such as sensors and other devices offers great promise -- including analytics that can be performed on the collected data.

• By Linda L. Briggs
• January 20, 2016

TDWI research indicates a growing interest in collecting and analyzing data from the Internet of Things (IoT). In the second part of our two-part interview, Dell's general manager for advanced analytics, John Thompson, explains some of the many benefits he sees ahead. "The value that we can get from analytics and the IoT is almost unfathomable at this point," Thompson says. (Read the first part of our interview here.)

BI This Week: We talked about the structure of a typical IoT network earlier in our interview. Regarding analytics, where does that typically take place?

John Thompson: There are many possibilities. [In one example scenario,] you might picture devices and sensors, with the data flowing through gateways into the cloud and back, into either a cloud data center for storage and computing or into an enterprise data center where there's storage and computing potential and where it can be backed up. In this scenario, devices and sensors are typically not a place where analytics is going to run.

However, given the way we've built edge analytics and capabilities [at Dell], if the sensors are capable enough, you can do some very simple filtering at the edge. Let's say you're interested in state change. If a light bulb says it's on a million times, you don't really care, but if it comes back and says, "Hey, I failed," that's a state change. You're interested in that and you want to put that state change up onto the network. That sort of simple filtering can reduce the amount of data flooding through the network.

Let's examine this a bit more. We've built analytics [at Dell] so that analytics can run in the gateway. That means you can have all of these sensors feeding into the gateway and you can have advanced analytics -- not just simple analytics, but very advanced analytics -- taking place right there in the gateway. If we picture a network that is a bit more sophisticated than what I described above, it would show another level of gateways or servers. Analytics could also be done there. You could have as many intermediate servers in these networks as is required to do data aggregation, or consolidation, or compression. ... With where we are today, analytics can truly be done at the edge of the network, at any intermediate level in the network, or at the core. It's up to you.

Also, keep in mind that as I said earlier, the core of one network could be the edge of another network. We've built the technology so that analytics can be done anywhere in the network -- you can have a truly intelligent network.

What are some examples of running analytics on IoT data?

With Dell [customers], we literally have hundreds of proofs of concept going on right now -- everything from environmental uses in buildings, to industrial applications, to financial uses, to health care.

Here's a great example from a proof of concept we ran early on when Dell was considering whether to get into the IoT realm. ... I think we all intrinsically felt that was true, but we wanted to go out and make sure it made sense to get involved with this initiative. We hooked up sensors to the boilers and elevators and physical plant at a hotel in San Diego. When we got the data back, we noticed almost immediately that the boilers were cycling in a very inefficient way. We did some analysis on it and came back to the hotel operations staff and said, "Look, if you change the way these boilers cycle to provide hot water throughout the hotel, you can save on energy costs." I believe making that change saved something like 30 percent of the hotel's energy expenses just in that one subsystem in that one hotel.

That was really a proof point for us. We said, we can look at efficiency in operational systems of all types and we can help make those far, far more efficient.

The implications for energy savings are immense. If you take that idea of efficiency and carry it to a Smart City initiative, you might be able to calculate how to cycle stoplights in a certain way to improve traffic flow and reduce gasoline consumption. In office buildings, if we hook up lights in a smart way and we have sensors around, we can detect if there are humans present. If you drive around any city in the world, you can see high-rise buildings with lots of lights on. Often, that's because someone walked out of the room without flipping the switch. Think of how much energy we could save if we could just sense that there's no one in the room and turn the light off.

What are the benefits of running analytics at various places in the network?

If we're putting analytics and filters all the way out to the sensors, as I said, one huge benefit is reducing the amount of data that flows across the network. People are talking about numbers approaching a trillion sensors on a network. A trillion sensors can generate a lot of data, but there's really no need to stack all that data up in the data center. Why keep five years of signal data from lightbulbs saying that they are on? That seems wasteful and not really sensible. Filtering the data at the edge is a benefit.

Another benefit is predictive maintenance. Think of the IoT networks I've described as more circles than just spokes -- as networks of gateways rather than just one. You can have many gateways and each of those gateways has the ability to speak to other gateways. Analytics can dynamically route the traffic and connectivity and say, "OK, we've found something that is probably going to fail. It's showing signs that in the next month or week it's going to fail." Many, many case studies show the amount of money saved by taking a piece of equipment out of service before it completely fails. It could be a diesel engine running part of a power grid; it could be a simple gateway, it could be an HVAC system in a retail environment. It's always better to trigger a warning before something fails.

Here's another example. You can have intelligent networks that are load balancing themselves, but the load balancing can allow you to predict when something is going to fail. The intelligent network can say, "This gateway is showing signs of failing." It has the capability to run at 50 percent of its normal abilities, so we're going to syphon off 50 percent of the traffic because we know that it could run for five months like this. We're going to send someone out in the next month to have it replaced, and then we'll re-balance the network at that point. That sort of predictive maintenance is a massive benefit of running analytics on the IoT.

Moving back up the network into the area of core analytics, we get into some really interesting benefits. We can make cities smarter or we can increase fuel efficiency. We can look at entire networks of buildings and the idea of turning their various systems off and on at the right times.... There's so much we can do with this.

The value that we can get from analytics and the IoT is almost unfathomable at this point. We can make buildings operate better, make cities operate better, and change the entire energy footprint of vast parts of our economy. There's great promise there.

As you talk to customers, what's the best place within an enterprise to try out an IoT network, especially along with advanced analytics?

Often, I find that there are areas where companies already have sensors and data -- they have information flowing through their organization already. We often start out with this: What are you doing today? We've been doing IoT-like analysis for an energy company since 2006, so this isn't something that needs to be green-field and brand new. You look at an enterprise and say, "OK, what kinds of business challenges are you faced with?" Often, there is an IoT data stream that is happening already. Then we have to ask, "What are you challenged with? What are your business objectives? What are the data flows that you already know exist? You can leverage those today."

If you don't have any of that today, then start with a green field. Maybe you're fortunate enough to be building something from scratch and you want to measure it in a new way and think about it in a different way. We always say, "Let's design with the business result in mind."

In my three decades of experience, it's always less effective to build something and then at the end say, "Hmmm, maybe we could do something with the data." Instead, design from the beginning with the data and what you could do with the data in mind. Look for existing environments that could be made more efficient, look for business opportunities and challenges that you already have, and look for businesses that are already running and have data flows. If you find that doesn't exist, then you can start from the very beginning.