Even if you have rudimentary estimating capabilities and you spend a little time doing back of the napkin calculations, you’ll quickly figure out that the notion that “all data is going to the cloud” is patently absurd. In the world of the industrial Internet of Things (IoT), data is collected “in the field.” And in the field, especially in remote locations and developing countries, there’s too much data and too little bandwidth at too high a price to move all data to the cloud. Even if you could, why would you want to?
Many in-the-field decisions, like “When should I replace this windmill part?” or “When should I shut down an oil rig pump to prevent a blow-out?” are based upon the results of rapid analysis of local data. Passing the data from the field to a cloud-based, big-data analytic app and then passing the answer back to the remote location could mean that the answer arrives far too late to prevent a disaster or a breach.
For these critical, real-time, decision-support applications, data analytics on relatively small data sets (often a few TBs) is best done at each location on highly available, but very low cost infrastructure. Other desirable attributes are low power consumption, small footprint, and tolerance for high temperatures and vibration. Most important is the ability to remotely deploy, monitor, manage, migrate, and perform software upgrades.
Increasingly these analytic apps are running on industry-standard IT infrastructure. Data for these higher-level applications may be collected from more proprietary systems common to industrial infrastructure of 10-20 years ago, but the emerging, more innovative applications are built on top of tools and platforms that have become standard and widespread in the IT industry.
Real-time monitoring and management apps based upon IT industry standard tools is becoming more and more common. Organizations just can’t afford to use the same datacenter solutions at the edge of the enterprise because of cost constraints. As these new-age remote sites are being designed and/or updated, its probably a good idea to consider a lightweight, software approach to compute and storage.