The devices, the connections, the platforms – each layer held together by a glue of interfaces. IS breaks it down into bite-size chunks
The new generation of IOT technology is a phenomenally powerful tool. It allows Digital Enablement – taking manual processes or decisions based on educated guesses and enriches them with hard information taken in real time directly from your physical environment – from machines, production systems, rooms, fields, vehicles.
Maybe you’re not ready for something as ambitious as a full-blown Digital Enablement programme just yet. No worries, you can still get your feet wet, because there are a thousand uses of IOT tech that can solve a small, but important, business problem. You can buy low cost devices off-the-shelf, and hook them up (securely!) to the Cloud in minutes. A low cost tracker that tells you if something is on or off, if it’s too hot, if it’s moved, if it's in a place it shouldn't be.
The Internet of Things is not a radical new technology – it’s an extension of the well proven techniques in wireless telemetry and “machine to machine (M2M)” systems. What has turned IOT into the hottest topic in tech in the past few years is the low cost and ease of deployment of new-generation IOT tech. This is because “Low Power Wide Area Network” (LPWAN) wireless network technologies enable devices that are very low cost (R250-R500 all-in), with equally cheap connectivity that costs a few dollars per year.
Why do you need IOT technology? Because it increases your visibility into the organisation’s day to day operation. They say you can’t manage what you can’t count – but you can’t count what you can’t see.
So – you want to implement some IOT. You need to build a three-layer cake, consisting of DEVICES, CONNECTIVITY, and DATA CONSUMPTION AND ANALYTICS.
IOT devices range from cheap and simple to elaborate multi-sensor smart devices – but it’s the cheap and simple ones that are often the most interesting.
Simple devices means cost is low for wide deployment, hardware is robust for “place and forget” and the application is clear. Intelligence and management is moved into the Cloud, where economies of scale of compute and storage can be bought to bear.
Simple IOT devices are usually designed at a hardware level around a particular communications technology. A Sigfox device is always a Sigfox device, a LoRa device is always a LoRa device, a WiFi device is always a WiFi device. Note that this is often, but not always, the case – some vendors have multi-radio devices, and this will get more common in time. But for now … an IOT device generally only speaks one “language”.
Because LPWAN technologies are incredibly power-efficient with heavily integrated chipsets, devices can be very small and battery powered. A battery that can last 5-10 years, together with the wireless LPWAN connectivity, means deploying a device can be as quick and simple as screwing it to the wall.
|Sensor Class||Description||Example Technologies|
|Simple Sensors||One (or few) discrete sensed values||Water flow|
|Complex Sensors||Several sensed values||Movement via vibration, GPS and accelerometer.|
Fire detection through air particles, temperature, light
|Part-time Sensors||Devices with sensing capabilities||Smartphone’s ambient light sensor|
Wi-Fi Access Point’s radio frequency sensing
|Synthetic Sensors||Multiple sensors’ outputs synthesised to detect secondary events||Ambient noise, EMI spikes and RF noise pattern to detect activation of domestic appliance|
Something to bear in mind is whether a device is “smart” or “dumb”. Irrespective of how many sensor inputs it has, does it have the capability to do any onboard processing? This could be analysing a visual feed to identify an event, or applying an algorithm to high-resolution inputs. “Smart” can also mean that the sensor manages its own communications – has its own IP network and security stack and connects itself to a back end. A “dumb” device would use a fixed connectivity configuration set up at manufacture or deployment time, like LoRa or Sigfox.
Obviously, a “simple” and “dumb” sensor will be radically cheaper to deploy than a “complex” and “smart” sensor – but will be much more limited in capability.
How sensors are connected depends almost entirely on your use case – there are a huge variety of technologies, but there is no one size fits all. It comes down to four elements: cost, power, range and data rate.
Data rate: is your device pumping out high resolution raw data in real time, or does it wake up to send a tiny, highly encoded message every few hours?
Range: does the device need to talk over tens of kilometres wirelessly, or can it be hooked up with a length of cable to a nearby controller?
Power: does your device have mains power, or is it meant for mobile or remote deployment and needs to eke out the milliwatts from a small battery?
Cost: do you intend on deploying a couple of devices where each costs R100 per month for connectivity, or do you need to deploy thousands of devices costing only R2 per month?
For example, Wi-Fi provides hundreds of megabits per second, but over only 100m or so (and doesn’t like obstacles), while being fairly power-hungry. Wi-Fi chipsets are extremely cheap, as is the access network equipment, and has negligible or no radio licence or carrier costs. Sigfox, on the other hand, will transmit over 10 or 20kms from a hole in the ground, uses minute amounts of power and costs fractions of a penny to send a message – but you’d have to be happy with only transmitting a few bytes every ten minutes at best.
|Sensor Network Type||Description||Example Technologies|
|Hard wired||Physical wiring looms connecting devices (for signal and/or power supply)||Multi-core copper wire carrying raw or encoded signals (e.g. Modbus, RS232)|
|Industrial wireless network||Wireless technologies designed for and used in industrial facilities||Zigbee|
|Wireless Local Area Network||General purpose short range wireless data networks||Wi-Fi (sometimes enhanced for industrial use)|
|Low Power Wide Area Network (LPWAN)||Low bandwidth, long range and low power consumption networks||Sigfox |
|Mobile Network||Mobile network technologies operated by national operators||GPRS (2G)|
|Satellite||Transmission via space-based transponders||R-BGAN|
Because the use case determines the best connectivity technology choice, it is all but inevitable that an organisation will need access to several types of networks.
Successful IOT implementations will be able to take advantage of a variety of device and network technologies, each with their strengths and weaknesses. This is where Internet Solutions provides a compelling advantage. From a single online portal, clients are able to procure connectivity across a number of LPWAN network operators, aggregate the traffic into a single place which maps between the various networks’ private addressing schemes, enrich and transform the data to expose vital metadata, and then reliably and flexibly route it to a variety of endpoints, supporting a variety of popular interfaces (including REST, MQTT, AMQP). No-one building the enterprise applications needs to know or care about the vagaries of the underlying IOT device or connectivity technology.
Data Consumption and Analytics
Once the devices are selected and installed, their connectivity is set up and they’re talking to the Internet – that’s where things get genuinely interesting. Something needs to be done with all this fantastic new data. Software applications have to be set up to listen to your Internet of Things – to give you the insights you crave, the operational visibility you need, the finger on the pulse of your day to day environment.
The Data Consumption elements can be loosely grouped into Visualisation Systems, Analytics Systems, Process Control Systems, Management Systems and Enterprise Resource Planning Systems. These can provide a range of functionalities, from simply showing you what's going on, to using data to make predictions about the future – and act on them.
|Sensor Network Type||Description||Example Technologies|
|Simple Visualisation||One (or few) parameters charted for real time and historical views||Time-series plotting tools|
“Traffic Light” tools
|Complex Visualisation||Multiple parameters from multiple sources charted with business logic visualisations||Building Management Systems|
|Intelligent Systems||Sensor data transformed and manipulated and intelligent analytics to produce insights||Production Management System integrated into ERP systems|
|Predictive Systems||Multiple data sources generating “Big Data” volumes driving Machine Learning systems to generate insights and predictions||Fully autonomous and self-learning systems|
Data Consumption obviously can’t happen if the data being fed to the applications is in the wrong format or not structured correctly. Data Consumption will inevitably need data transformation layers.
Finally, a crucial factor in Data Consumption is the sensitivity and operational criticality of the data. A consumption system may be very basic – but if it is showing parameters impacting on human life, it has severe governance requirements. Fail-to-safe, multi-level alarming, non-tamperable archiving, etc.
This may all seem a little overwhelming – the number of choices that can be made, and the fact that there may be unexpected technical limitations, or integration complexities – never mind long term information architecture, security, governance and even compliance issues.
The good news is that modern IOT technologies are designed to be flexible, and to adhere to common open standards for information exchange. Most complications can be worked around after the fact – and with the connectivity options being very affordable on public LPWAN networks, it’s not a massive issue to run multiple technologies.
As Internet Solutions, our recommendation is to start at either end, and meet in the middle of the cake.
Device selection is absolutely driven by the specific use case, and which products are available in the market (or can be feasibly custom-developed). Information architecture, software stack and long term IT strategy and governance are also known quantities, defined by the enterprise. With these two layers known, the connectivity choices can then be made to stitch the “cyber” and the “physical” worlds together and complete the cake.