Aug 28, 2017 Get Ready to Bring the Data—All of the Data
PARTNERS Session: How to Implement Industrial IoT at Scale
“Bring data to the analytics, not analytics to the data.” That’s Artur Borycki’s, from Technology & Innovation Office at Teradata Labs, advice to those thinking, designing, and implementing IoT (Internet of Things) solutions. “Don’t be scared to try the new approach. Don’t be bounded by your existing procedures and the way you approach the problems. Even at the prototype scale, you have to bring all of those different variables so that you can correlate much larger volumes of the data.” That’s a new way of thinking for a simple reason: it previously wasn’t possible.
As Artur says, “We’ve been doing IoT for the past 20 years, it just wasn’t called IoT. Today we have access to more data, and finally have enough technology to start to operationalize those insights. It’s exciting because we’ve finally reached the point where we can look at all of the data, all of the time. We’re not bound by data volumes, or computing power. We can start correlating information to generate more value.” These new capabilities will have widespread impact both in the results that we can achieve but also how those working to solve the problems need to approach them.
Bigger Data Sets for Better Quality
That change has made itself apparent quickly in complex manufacturing. For example, if you are striving to increase the quality of a product, the tricky part is to find what is causing the problems because you might have to look at the millions of variables per device. “It was very hard to say ‘this particular part of the machine is causing the problem,’” according to Artur. “You need to find the correlation between those problems, and then identify the clusters of those problems.” In the past, people were limited to looking at a subset of the data because of the large number of variables per device. That’s not the case anymore. “Today they can look at all of the data, all of the variables, and say this specifically shows the correlations, and then improve the quality of the product.”
“Their problem wasn’t the use case, it was the scale. They were limited by the size of memory to which they could load the data.”
– Artur Borycki, Technology & Innovation Office, Teradata Labs
Think Bigger from the Start
Starting with the end in mind is good advice for any project, especially IoT. “Even when you’re experimenting, keep in mind that to move this to production will require you to analyze bigger volumes,” says Artur. “In the industrial IoT space we are not talking about the small volumes; we’re talking about the big problems and the large volumes. You have to understand the scale of the solution you are going to provide.”
For example, Artur was working with a data science team that came to him saying that they’re capable of finding the quality correlation, but they can only do it at 10,000 rows. “Because they are thinking thousands of variables per row, they really cannot check if this is the only correlation, or if there are some others which are related and can make this more relevant. Their problem wasn’t the use case, it was the scale. They were limited by the size of memory to which they could load the data.”
End Goal: A Happy City
Ultimately IoT will be everywhere. That includes where we live, but beyond the smart home you’ve heard so much about— into the cities and beyond. This is an area where Artur is especially excited. “Right now a lot of cities are focused on optimizing their overall costs. They’re looking at how I can balance power utilization for example, because this is their biggest problem at the moment. But it can be more than that, what I’m calling the smart nation. That’s where I can optimize not only the costs of living in the city, but basically make people more happy, by understanding how people are moving around the city, and therefore provide them better information, more efficiency, and less stress. This could be anything from commute routes to hospital wait times.”
You’ll Hear It Straight at PARTNERS
If you haven’t been to PARTNERS, you should realize that it is more open than other conferences. “The great thing is you have customers who are not afraid to talk honestly about what they are doing, and what is their experience,” says Artur. “This is always an amazing experience. I’m looking forward to seeing what customers are doing, what are the pitfalls they are facing, and how are they using the technologies.”
Don’t miss Artur’s session or the ten other sessions on the subject of IoT. Make plans to register and attend today.
How to Implement Industrial IoT at Scale – Session 0167
Internet of Things is becoming widely discussed not only in the case of sensor data, but at the same time the requirements are demanding more scientific analytics at scale—we need to look to several ways how this should be architected. This session will present several examples of how Teradata and customers have implemented IoT-enabled architecture to support all of the requirements from real-time ingestion and communication with Edge systems to analytics at scale. We will also include how we can train and deploy analytical models back to the end user or device.
Technology & Innovation Office, Teradata Labs, Teradata