Architecting the Industrial Internet

With the Industrial Internet, big control meets big data at the convergence of machines, intelligent data, and people at work. Machine control, big data analytics, resilience and security cannot be an afterthought — it has to be part of the actual fabric of the architecture for the Industrial Internet.


The graphic illustrates the interplay of machines, data, and people — alongside control and security. Machines, as well as enterprise and exogenous systems, amass large volumes of data, all the time. Data is processed (on machines, on premise, in the cloud), analyzed with predictive and other models, complemented by advanced asset management systems. Actionable information that is shared with people at work, and (via embedded analytics) processes automation software. Closed loop circles take insights, feed analytics, and control actions of machines.

Given this environment, a software platform for the Industrial Internet must fulfill four requirements:

  1. It needs to be machine-centric.
  2. It needs to be optimized for industrial big data.
  3. It needs to follow a modern architecture paradigm.
  4. It needs to be resilient and secure.


The Industrial Internet begins (and often ends) with machines. Machines generate increasingly large amounts of data via increasingly sophisticated yet affordable sensors. And machines take actions based on insights generated from often real-time data analytics, via controllers and actuators.

Increasingly, machines need to be connected to provide discrete services such as basic machine identity to performance reporting. Machines need to communicate with plants and fleets, as well as with one another, often in real time, with guaranteed response times, over multitudes of wireless and wired networks. Machines may need to locally process large volumes of high-velocity data (for example, an aircraft engine). And they may connect to data centers or cloud infrastructures to process large-scale predictive data models across entire systems of machines (for example, flight route optimization).

The Industrial Internet requires a software platform that provides a standard machine-app model to build, port, and deploy those machine-centric Industrial Internet applications at scale.

Industrial Big Data

Industrial big data outshines other big data in terms of volume, velocity, and variety. Data growth for machine generated data is more than double that of any other data. Velocity of data generated is higher than in the consumer Internet. Variety of sensors and machines is more complex, too.

A software platform for the Industrial Internet needs to provide an order of magnitude improvement in the scope and quality of big data analytics and operations. And it needs to do so across verticals: The real-time analysis of a jet engine in flight has different requirements, data quantities, and use-cases than the real-time analysis of health care information.

Big data processing needs to get coordinated across machines, server farms and clouds, over reliable high-speed networks. Distributed analytics need to scale-out elastically, and scale-in as well to the individual machines’ micro-controllers. Recent big data paradigms (such as Mapreduce / Hadoop) challenged older business intelligence approaches, by bringing the processing to the data (versus the data to the processing, as in the past). With the Industrial Internet, the approach needs to be taken even farther, by bringing the processing to the machines themselves.

Assets are the "master data" of the Industrial Internet. Not unlike to classic master data management approaches, management of these assets needs to be tightly married with big data analytics processing. The software platform for the Industrial Internet needs to excel in both — big data analytics and big data asset management.

Modern Architecture

With the rapid evolution of the Industrial Internet, OT (operational technology) and IT (information technology) professionals need to work hand-in-hand to rapidly deliver Industrial Internet apps. This requires a service-enabled "cloud-agnostic" platform, to enable the rapid design, development, and deployment of business-relevant apps. Similarly, that platform should be cloud agnostic, to deliver on various industrial, regional, and corporate mandates — supporting the deployment of applications in the private or public cloud, on premise, or all of the above (hybrid cloud).

Along with new software deployment and delivery models comes the need to refresh the user experience in the industrial world and deliver advanced insights (e.g., over mobile devices). Aligning the industrial user experience with the consumer-grade experience not only will help drive a broader use base for industrial applications, it will allow industrial companies, many of which have an aging workforce that will soon retire, to create a user environment that appeals to a new generation of workers.

Extensibility and integration are critical, too. A platform needs to support the integration of existing enterprise IT technologies, platforms, devices, and software into the Industrial Internet. So is the ability to better integrate the enterprise back office with industrial operations: Once the machine predicts a material failure, a replenishment process is started to backfill parts into inventory.

Resilient and Secure

The enterprise world traditionally looked at system uptime in terms of five or seven "nines" of availability (99.999% to 99.99999%). These service levels may be unacceptably low for many industrial applications: when the safety of an airplane or nuclear plant is on the line, much more than seven "nines" of service availability may be needed.

How devices are connected to the Industrial Internet, provisioned, and decommissioned must also take place under a much more stringent set of requirements. Connected devices need to be instantly on and instantly credentialed and authenticated, as opposed to existing models of discovery that require a process that may even include an end user or technician physically inputting an upgrade.

Last but certainly not least, a software platform for the Industrial Internet needs to support a wide variety of safety and security regimes, often mandated by regulations.

In summary, a software platform for the Industrial Internet needs to be machine-centric, optimized for industrial big data, built on a state-of-the-art modern architecture, and deliver on the most demanding resilience and security demands of the industries.