The Internet of Things (IoT) is gaining steam as businesses across various industries launch projects that instrument, gather, and analyze data to extract value from various connected devices. While the general vision for IoT may be same - each company is pursuing its own unique approach on how to go about it. The adoption of standards and emergence of industry leaders will help the “wild west” situation we’re in but it is still unknown how long it will take to get there. How should businesses implement their IoT solutions in a way that will allow them flexibility and control no matter what the eventual IoT landscape looks like?
It is relatively easy to put together an IoT solution using
consumer grade components. The companies catering to the IoT market are coming up with ever cooler devices that make it easy to gather environmental data and send it to the cloud. The proof of concepts (PoCs) created by lines of business, marketing, or the CTO office often use these consumer-grade components to put together their IoT solutions. That said, to deploy industrial or enterprise-grade IoT at scale within a company, the challenges are going to be far different than what it takes to put together a PoC.
Enterprise IoT will typically encounter some of these physical and technical contexts:
- Oil & Gas: Extreme work environments with intense temperatures, contaminants, and vibrations. The scale of the deployment is massive: there can be over 30,000 sensors/actuators on a single oil rig. Some of these devices are pretty expensive (e.g. a quartz-crystal pressure sensor used in an oil well can cost up to $20,000).
- Manufacturing: The average lifetime of factory equipment is 10-30 years. These systems are managed by operations technology groups. Any new IoT system will need to integrate with existing factory systems without compromising factory safety. These IoT systems also need to be maintained, patched, and updated over very long (decade-long!) lifecycles.
- Smart Cities: One of the many forms of IoT - smart parking - is becoming popular with cities using technology to help the environment and citizens, and as a side effect, generating greater revenue. It benefits citizens by minimizing the time it takes to find parking space (and pollution, cost) by matching them to the closest parking space. Cities can generate greater revenue by automating parking compliance and also use surge pricing to dynamically configure pricing based on demand. However, the parking systems need to have fail-safe capabilities so they can continue to function even in a degraded capacity.
- Hospitals: Automation can improve patient care by reducing caregiver mistakes, for example a smart drug infusion pump can reduce costs by improving asset utilization and create a more hospitable (no pun intended!) atmosphere for patients through customized care. These are some of the top priorities for hospital management that can be achieved through enterprise-grade IoT systems. However, these critical systems are not as secure as desired. Take, for example, the FDA warning on the security hack on Hospira Symbiq pump and the subsequent recall of the entire line of these pumps (compounded by other software glitches).
It is clear that the environment for enterprise IoT dictates far more complex requirements than what it takes to build a PoC. The approach used to put together a consumer IoT solution using fully integrated albeit proprietary solutions (e.g. ThingWorx) or built-from-scratch solutions using DIY technologies (e.g. Raspberry Pi) are not going to work for enterprise-grade IoT implementations. In addition, the PoCs using consumer components are not going to be integrated with the company’s existing IT infrastructure and will operate in their own silo requiring them to have their own database, network, security, apps, and analytics. They can be quick to assemble but are designed for one-time use and lack the rigor of production quality testing, security policies, maintenance planning, and ongoing support.
Also, worth mentioning is that these PoCs typically have a 2-tier architecture where data acquisition is at the edge but where processing/analytics and intelligence are in the cloud. The cloud service providers (e.g. AWS and IBM Bluemix) make it very easy to connect to their respective services by providing pre-configured deployables and APIs that can be connected to various data sources. However, their pricing model is more suitable for PoCs as costs can rack up pretty steeply when deployments scale beyond the PoC.
Let’s look at the scale of data being generated for industrial IoT:
- Aircraft engines generate 1-2 TB of sensor data per flight (3).
- Refineries generate 1 TB of data per day.
- Locomotive engines generate 216 million data points per day (2).
The cost of sending locomotive data to the cloud (1) would cost over $1,000 per day or $400,000 per year. This doesn’t include the the transmission costs that could be very expensive over cellular networks! Nor does it include the cost for messages sent from the cloud to end devices or compute resources to process this data.
As much promise as IoT holds, the full-scale adoption rate has been lower than analysts’ forecasts. Very seldom do the PoCs reach the scale to make any real difference beyond the group that sponsored them. The most likely causes could be the lack of testing, integration, maintenance, and support needed for a production quality deployment. Also, the 2-tier architecture where data flows from device to cloud is not economically feasible for businesses at a mass scale. In my next blog, I'll explore how to overcome this issue by bringing intelligence closer to edge. In the mean time - if you have thoughts or questions - I encourage you to reach out using the comments section (below).
2 - http://www.cnet.com/news/at-ge-making-the-most-advanced-locomotives-in-history