The Internet of Things stands for the networking of intelligent devices and machines. But what exactly has to be networked and what IT architecture is necessary to process these huge data streams?
In 1988, the US scientist Mark Weiser coined the term “ubiquitous computing”. This refers to a computer ubiquitous presence. Behind this term lies the trend that more and more everyday objects and machines contain powerful IT components that connect to each other via the Internet or interact with each other.
The Internet of Things (IoT) is regarded as an important technology trend. Market researchers at IoT Analytics Research expect global sales in this sector to exceed 146 billion US dollars in 2018, which could rise to over 357 billion US dollars by 2021. In addition, there should be about four IoT devices per person in 2020, i.e. a total of 20 billion networked devices. Growth rates are expected to remain high.
The management of such an immense number of devices requires a powerful IT architecture and infrastructure. Without the use of cloud services with scalable services, it is practically impossible to master the huge amount of data generated by the IoT.
Sensors are covering the world
As the development of the last decades shows, more and more sensors connect the digital with the analog world. As early as the 1990s, radio tags based on RFID technology (Radio Frequency Identification) were used in toll collection systems in the USA, and since the turn of the millennium RFID has been used for goods identification in retail. The trend to install more and more sensors in everyday objects – now, however, integrated via the Internet – came up in the 2010s. This also increased the need to collect, process and permanently store data obtained from sensors. Networked sensor systems can now be found in more and more areas, from automated factories to self-propelled vehicles. At the same time, cloud providers are developing powerful platforms for managing IoT data flows and evaluating sensor values.
Things and their data
The “things” themselves form the basis of the Internet of Things, from simple sensors such as temperature sensors to complex machines with comprehensive sensor technology. The tasks of these devices are correspondingly very different – from less complex to very complex. It could be a microcontroller, such as a Raspberry Pi with only one rain sensor. But a dump truck with hundreds of sensors can also be defined as an IoT device that transmits vehicle data several times a second via an interface. A complex commercial vehicle supplies a wide range of data, such as the current tonnage of the load, the tipper’s angle of inclination, and status values for the engine, transmission and cooling system, right up to the current speed.
This example illustrates the amount of data generated by a vehicle alone. Here is a small calculation example: With 1,000 sensors that deliver measurement data of 8 bytes 10 times per second, a data volume of 6.44 GB is generated per day. This corresponds approximately to the data volume of one hour of film in HD quality. If a company owns more than one tipper truck and these vehicles are then used in underground mining, data processing must first be carried out on site, as no mobile radio network is available. In this scenario, Edge Computing should be used, which utilizes IT resources directly at the customer’s premises to preprocess the data and transfer it to the cloud. With the IoT, IT infrastructures are also becoming more complex. In contrast, a small Raspberry PI computer, which only determines the current temperature once per second, can also transfer this data directly to the cloud.
Given the diverse requirements of an IoT architecture, experts are needed for data engineering. Another look underground: filling the tipper will take several minutes, so that 100 measured values per second do not have to be transmitted here. Furthermore, sensors that are only supposed to return error values, for example, often always return the value “0” for hours or days (but transmitting “0” costs of course data bandwidth). It is therefore necessary to remove the “noise” from the data and to bring it into an adequate form. Only the important data should be transferred to the cloud, but none should be discarded that may be needed again later. From a technological point of view, the transfer to the cloud takes place via a gateway, whereby networks such as WLAN, LoRaWAN (Long Range Wide Area Network) or UMTS can be used.
It is therefore important for companies to approach the IoT in particular via the data topic. In this context, various aspects need to be clarified, including which data must be processed in which form, which data is relevant for operational business or how data can be processed at the point of data acquisition. Among other things, the network routes and data storage have to be considered in order to consider topics such as data protection, compliance, network management or quality of service.
At a glance: How the Internet of Things works
There are several IoT infrastructure approaches, but no universal solution for all scenarios. Companies need an architecture to collect and process data in a meaningful way. Based on this, it is possible to set up a business logic or a graphical evaluation.
Such an infrastructure can be set up successively. In the first step, it is possible to move the data unfiltered into the cloud and process it there. If an initial data analysis or filtering on site becomes necessary at a later date, edge capacities can be retrofitted there, for example on the basis of AWS Greengrass. In principle, as much pre-processing as possible should take place “on the edge”, i.e. on site, in order to save costs and time.
We at Storm Reply support companies in the development of an IoT business case and advises on the selection of the right data. With a prototype or minimum viable product, our experts show how data can be transferred to the cloud or data visualization can be realized and which findings companies gain from this. In addition, Storm Reply also advises on further steps up to the continuous expansion of a comprehensive IoT platform.
In the next two blog posts, we present two different approaches for an IoT architecture.