Machine Learning Cuts the Industrial IOT Value Delivery Problem

by Simon Crosby, on Aug 23, 2017 11:56:52 AM

The concept of an IOT application platform is already an anachronism. Scan the web and you’ll find over 400 vendors, from startups to huge firms such as GE making the case that their platforms are the future of edge analytics for IOT. Their narrative has roots in the fabulous innovation of the last 10 years by the major cloud vendors, starting with the concept of Infrastructure as a Service (IaaS), and progressing (at least theoretically) to Platform as a Service (PaaS). Of course Software as a Service (SaaS) is a huge winner here, spanning everything from Office 365 to All in all, there is a general perception that the cloud solves all problems, and so when it comes to IOT analytics, and the need to support a diverse set of applications and edge scenarios, then the cloud narrative is a no-brainer:

This narrative says that success depends on how well you build, orchestrate and run your new application in the cloud. There are thousands of open source stack components available, and success is merely a step away from picking the right PaaS stack (and perhaps an embedded protocol or two), and some development.

This narrative leads to failure. The success and importance of the cloud cannot be doubted: Enterprise infrastructure should move to the cloud as quickly as possible simply because owning and operating computing infrastructure is no longer of value in delivering IT services. Similarly, wherever appropriate, enterprises should consume SaaS applications, rather than implement their own solutions on premise.

But in the Industrial IOT and Smart City world of edge analytics, there’s a huge problem. First, industrial automation solutions (what IIOT was called before it became sexy) have tended to be delivered in a combination of equipment and software, integrated by a major vendor. Breaking functionality out of a box at the edge and replacing it with a cloud-centric notion of the same function just doesn’t make sense: Local decisions based on real-time data need to be made locally, and not in the cloud. But even on a slower time-scale enterprise and equipment vendors lack the dev skills to build new apps in the cloud, IT/OT can’t manage, secure or scale cloud-based applications, and getting access to data from brownfield deployments that can’t be replaced is difficult.

Luckily Industrial IOT is not an App nor DevOps problem. It’s a data problem.

Better still, it’s a data problem that we can tackle with machine learning.

The “app platform” approach has its roots in the “big-data, public cloud IaaS solves all problems” narrative. But instead of slaving away at new apps, we at Swim have found that we can deliver immediate value to our customers with almost no domain knowledge, by simply learning on their data, on-the-fly, in context, at the edge. Learning on cheap devices close to where the data is produced has many benefits:

  • We can learn on brownfield data to short-cut app building, without the need to rip and replace.
  • Using cheap devices, we can learn locally to deliver immediate insights and quick response
  • We can learn on commodity devices to cut bandwidth, storage & cloud processing costs
  • Learning simplifies adoption of new solutions, overcoming the IT/OT skill set problem.
  • Learning through introspection helps to ensure security and availability.
  • Machine learning can short-cut the IIOT value-delivery problem without the need for cloud skills or data scientists. That’s exciting!

Learn More

Learn more about SWIM.

Topics:Machine LearningSWIM SoftwareIndustrial IOTSmart CitiesEdge Analytics