Analytics and the cloud: The Internet of Things

Managing the growth of a connected world to gain new insights

Sensors are driving the explosion of data

Stop for a moment to consider where sensors act all around you. People can run entire businesses from their phones, vehicles are filled with diagnostic and climate control devices, homes have smart utility controls and meters, factories are becoming sensor-enabled as they automate, airplanes have a vast array of sensors to monitor flight performance, communities have sensors to manage traffic flow when road lights switch on. The list is already endless.

Data collected from connected devices is increasingly integrated with analytics to tackle a variety of problems such as:

• Home Health Monitoring – track patient’s vitals and their movements around the home, and raise alerts to family/physicians/hospital if abnormal events take place.
• Energy Monitoring – at a macro level, utilities can improve national grid effectiveness and power generation; at a micro level, home devices can be powered down or placed on standby when not in use.
• Asset Management – monitor the health of devices in factories or difficult-to-service locations (i.e. mines, remote wind farms); analyse data and provide diagnostics on devices to predict failure and build predictive maintenance schedules accordingly.
• Safer Driving – connect vehicles to understand road traffic, monitor performance of both car and driver, to offer lower insurance premiums for safe drivers, drive innovation toward more efficient and even driverless cars.
• Logistics – track packages and containers, provide estimated arrival times to customers, identify staff distribution centers where workloads are greatest
• Environmental – leverage weather and air quality sensors to predict when pollution may cause issues at city hotspots, monitor changes in water levels, localise and track data to finer grained geographical areas for greater accuracy in local forecasts
• Building IoT solutions to enable analytics – When building solutions to use all the data available there are multiple considerations, that together, introduce unique issues with IoT solutions.
• Velocity – Data is normally moving in real time and needs analysing as such. – Streaming solutions to identify changes to data patterns can be used in feedback loops to change the operation of devices. For example, a switch could be activated to release pressure, based upon a sensor’s data that starts to show untypical readings or pressure fluctuations. If done in real time, this can potentially prevent a serious fault in a gas or oil pipeline.
• Volume – IoT solutions may bring a variety of different data together from video, audio, vibration, temperature, pressure, humidity and more. All of these data sources need to be managed and derived data must be analysed appropriately. This can result in huge amounts of data processing, a genuine Big Data use case.
• Location – Sensors can be anywhere, and are often mobile. Networks can be complex and their capabilities need to be carefully considered. At the edge of the network where instability may occur due to environmental issues, fault tolerance needs building in. The need to aggregate data or build analytics close to the edge to reduce the load at any central point could be called for. This infers building intelligence close to sensors to either in dedicated devices (think of a raspberry Pi) or within the devices themselves.
• Standards – Currently there are many ways that data sensors can communicate to a central point. Bluetooth or Bluetooth Low Energy, Wi-Fi or 2G,3G,4G are all common. All are useful and need to be managed across an IoT solution. It may be that Blue Tooth Low Energy (BTLE) is used at the edge to connect multiple devices to an interim ‘aggregator’ and then that device uses Wi-Fi to send data onto the central systems.
• Cloud – The core systems that are used to analyse IoT derived data are located in the cloud. The demand for resources will vary over time and the processing requirements to model, predict, simulate and visualise the stored history can all be managed from a single point.
• Integration – IoT solutions often need much more than just sensor based data to become useful. Reports from engineers held in content management systems, ERP (scheduling), and Asset Management systems all within the enterprise may also need to be brought in. This requires integration across all system components.
• Security and Privacy – Maintaining the security of a set of devices that are distributed geographically, potentially using differing standards and networks is not easy. Being able to prove that such solutions and private and secure where they need to be may be the hardest problem in any IoT solution to solve.

The IoT landscape has some very interesting challenges indeed. There are potentially millions of sensors that must be managed with intelligence built in to control flow of data, local analytics is needed close to devices themselves, error management and feedback to physically control the ‘things’ the devices monitor if the devices themselves are capable of such are a few of the opportunities when we build an IoT solution.

IoT analytics in the cloud

All the data resulting from the output of these devices should flow into a cloud based solution for analysis. Thus, the cloud becomes the “brains” of an IoT solution. The cloud enables action to be taken with the gathered data including simple reporting, predictive modelling, simulation of differing outcomes to test hypotheses, and closed loop systems that communicate with devices to stop failing processes and/or automatically fix problems.

The cloud holds all historical information from the sensors, and can integrate it with data from traditional systems such as ERP, asset management and other forms of data. An IoT solution embodies all forms of systems—Systems of Record, Systems of Engagement, Systems of Automation—to deliver insight to the business.

What does an IoT architecture look like?

I’ve drawn upon work already completed by the IBM cloud architecture team to show the principle components of a IoT Reference Architecture. Primary areas of the architecture are:

• User Layer – the primary actors in the system, whether a user or application makes use of the IoT system.
• Proximity Network – the layer of the architecture that manages the sensors themselves, and provides connection to the things they monitor and with the public network.
• Public Network – links devices to provider cloud services. Other services to augment the sensor data could be fed in here, including data from government open data sources, weather data providers, social media etc.
• Provider Cloud – the layer that holds the core services that make use of data gathered from the previous layer. Here is where data can be transformed and analysed to offer insights.
• Enterprise Network – the final part of the solution is the existing data location within the business’s enterprise, for example ERP or asset management.

In conclusion

IoT has opened up a new range of analytics to manage data from a multiplicity of sources that 10 years ago wouldn’t have been considered. The advent of low power sensors, and their ease of use coupled with having lifetimes of several years before replacement, means nearly anything we consider will one day be instrumented. The continual upsurge in data and the need to make sense of it all creates demand for storage and systems, and resulting costs, that can be managed best in the cloud.

Only by linking data that we might not consider will we truly understand more about the variables that prevent processes from working as optimally as is possible. Analytics sited on the cloud is the ideal platform. It is on demand and can be operated as a managed service with continual development and easy roll-out of new services. This allows new ways of processing data to be brought to bear. Platforms such as Spark will enable developers to test out new ideas quickly and share them with others in a collaborative manner.

The IoT market continues to evolve to meet the demands of growth, but as it matures in the next 3-5 years’ leaders and followers will emerge. The leaders will not only have to be able to handle the complexity of connecting literally millions of devices in a secure, managed manner, they must also be capable of analysing all of the assembled data in real time and historically, whilst merging additional data from other cloud-based and enterprise sources.

IBM is well positioned to do this with the legacy of middleware for connectivity, established cloud based services, a rich history of analytical solutions, security platforms that are world leading, and our partnerships with other like-minded companies.