IIoT is the Natural Progression of your PI Environment

For those of us working in the PI world for the past couple of decades, we were Industry 3.5. We were the ones who broke up the silos and made process data available across the enterprise. We bridged the divide between the Information Technology world (IT) and the Operational Technology World (OT). By installing interfaces collecting data from the company control systems and pushing it through a firewall onto the IT network where everyone could consume the data, we brought increased visibility to core business processes allowing engineers, analysts, and management to scrutinize processes, optimize them, and report on production, efficiency, and environmental compliance.

The PI system changed the world for the better, making information from disparate control systems available to everyone and putting powerful tools in the hands of users. This is why thousands of PI systems are deployed in most Fortune 500 companies, and billions of PI tags are deployed worldwide. It has been quite a ride. After a couple of decades+ in the industry, it is only becoming more important and dynamic. I hope to explain how what you already know from the PI world applies to new trends and how Industrial IoT (IIoT) is a natural progression of what we already do.

The figure below depicts the bread and butter of PI. Industrial Control Systems (ICS) control important processes in our companies. In most cases these days, we configure an OPC interface to collect data from these control systems. The interface runs on a Wintel platform and translates the OPC protocol into PI Net to send to the PI server. It is a data conduit that ships the data, unaltered, from the source to the destination safely for mass consumption. Importantly, the interface node buffers the data during connection interruptions to the PI server, so no data is lost.

I will now introduce an IIoT platform, the open-source Linux Foundation project called Fledge, and show it doing the same job. I hope to talk a lot about Fledge in the coming months. Here is a little of my background. Fledge adds a few desirable traits to the equation. For starters, it runs on Linux, so you can deploy it anywhere that you can deploy Linux. Raspberry PIs, embedded platforms, Docker containers, Kubernetes, etc. It is a lightweight, modern architecture based on microservices. Most importantly, it is open source so you can modify and extend it. It talks about of the box to the PI system and it will collect data from OPC UA, which means it will talk to just about anything.

Like a PI interface, it will also talk directly to devices like PLCs using native Fieldbus. Modbus, DNP3, Ethernet/IP, and others are available. You don't have to use OPC.

Why would I choose to use Fledge or any other IIoT framework? Why not stick with what I know?  First off, your PI interfaces are only passive conduits for information. OSIsoft played with StreamInsight some time ago as a way to modify and filter information as it flowed through your interface but abandoned the effort. Many use cases make sense to process the data as it flows through the conduit. Vibration data is a good example. Normally, you might want to send RMS data to the PI server, but when an event occurs, you want to collect higher-resolution data and forward it for analysis. You don't want to send that high-resolution data all the time. This is just one possible use case. You can probably think of many others.

Fledge is like a PI interface on steroids.

  • You can filter and process the data as it runs through the conduit
  • You can calculate roll-up data, windowed aggregates, etc.
  • You can conditionally forward information
  • You can collect and forward non-time-series data, which can help train ML models
  • You can run ML algorithms against the data stream
  • You can send the data stream to more than one destination
  • You can collect information from many sources, not just the control systems

This is Industry 4.0, the logical progression of what we have been doing for years. As enterprise networks extend into the cloud, we gain access to resources we could not have imagined years ago. Still, we also realize that processing on-prem, at the edge, will be a critical component of this expansion. We need open frameworks to accomplish this.

  • Linux-based
  • Open source and governed by the Linux Foundation
    • Having an established and proven governance model will be critical
  • Extensible and plugin-based
  • Dockerizable, so you can still run it on a standard Windows computer
  • It runs in WSL under Windows if that is what you would like


This promises to be a wild ride. We are at a crossroads as the traditional ways of doing things meet their limitations, and new ways must emerge. We must all adapt and look for new tools to solve the problems we have encountered on our journey. Our knowledge and experience gained from years of supporting some of the biggest companies in the world are valuable resources on this journey. I hope that you will choose to follow me and contribute as I think out loud. I invite your feedback and criticisms as I document my thoughts in the hopes that we all learn something. As a wise man once said, "If you want to do something right, do it wrong and post it on the Internet." This is outside my comfort zone, but I think it will be worth it, for all of us.

In the coming months, I will post more ideas about the PI system and the improvements I have in mind. I will also post about how I think it is time for the industrial community to embrace open-source to spur innovation. I hope you will follow me and, most importantly, provide feedback. I've learned so much from our community over the years. I hope to give back a little and improve.

To see my posts about PI:


A little about my role on a Linux Foundation project that I hope will help change the world:


This blog:


My twitter feed. I think about this a Foley's after conference hours. Let's have some fun:




Schedule a meeting or call with me:



Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.