The Internet of Things (IoT) is a broad term. In general, though, it means machines talking to machines with a back-end system gathering the data from those conversations for analysis and control purposes. In the manufacturing world, IoT techniques and technologies can be the key to greater productivity and efficiency. That is, they can be, if you get them right.
How to get it right
Getting it right starts with figuring out what you want from the industrial IoT (IIoT).
- Will it be enough to take data from sensors so that your back-end analysis will let you begin predictive maintenance? For some companies, that alone will justify the project.
- Will you be using the data, again coupled with your back-end servers, for command and control based on real-time data?
- Are you going to set up a system in which your equipment is talking to itself, adjusting operations far more quickly than any human operators could?
The answer to these basic questions will inform decisions on everything from network topology to the servers required for analysis.
Servers and communications
Regardless of your other decisions, you’re going to need communications and servers for analysis. Both come together at the rack where your analytics are homed.
On the communications side, this means a network that is designed for high connection and transaction counts rather than for the largest possible raw bandwidth. IoT applications tend to involve incredibly high numbers of very small transactions. It’s easy to think that, for example, all gigabit Ethernet or 10 gigabit Ethernet network cards are created equal, but that is far from the case. Ask your vendor about the transaction and connection capabilities of the cards you intend to put into your IoT infrastructure.
When you begin to put together the specifications for an IoT infrastructure, you’ll want to look at the possibilities of twin servers for your IoT. One of the servers will be designated to receive and send IoT traffic, while the other will be the heart of the analytics process. The difference between them? Transactions.
The Industrial IoT is a classic online transaction processing (OLTP) application. The qualities needed in the storage design and overall application infrastructure are almost identical to those that would be used in a traditional centralized point-of-sale or e-commerce system. Key to acceptable performance is write speed, the ability to catch a fire-hose stream of data and write it to storage without pause or bobble.
Continuing to look back at classic processing, the analysis system will be a traditional online analytical processing (OLAP) system in which data writes can be leisurely while data reads need to be exceptional, delivering data at near-streaming speeds for accurate, timely analysis.
The Industrial IoT can put a mid-sized manufacturer onto a level playing field with much larger competitors. Build your infrastructure properly, and it will maximize the benefits while keeping the additional load on your IT group well in hand.