Cloud

Internet of Things (IoT) in global public clouds - six lessons learnt

The Swisscom Global Public Cloud AWS Tiger team has been working intensively on the Industrial Internet of Things. We have built an IIoT PoC to prove the feasibility of some key technical challenges for a future platform that has ten million machines and multiple geographic locations as a prerequisite. We also need to be able to scale with hundreds of thousands of users and tens of thousands of manufacturers of things. We've been looking at the OPC UA open standard, edge gateways and analytics streams. We have learnt a few lessons for the new manufacturing with IIoT in the global public clouds that we want to share.

By Wuming Zhang, Cloud Architect

An overview of the IoT

The Internet of Things (IoT) has been around for a while, the term was invented back in 1999 when RFID was first advertised. You may be thinking of Amazon Echo, Google Nest or Philips Hue. A Fitbit or a wearable is also a form of IoT. We had the opportunity to take a closer look at another type of IoT. Industrial IoT - a subset of IoT that focuses specifically on industrial applications such as manufacturing or agriculture. It is mainly concerned with two areas: Increasing efficiency and improving health and safety. One example of the potential of IIoT is predictive maintenance. A broken machine in a manufacturing process can mean millions of dollars in lost productivity while production comes to a standstill to fix the problem.

Every IoT solution consists of three parts. The edge level with things that recognise and act. The platform tier, which collects, stores and processes data and manages the IoT devices. The last part is the enterprise tier with intelligence based on data analysis. This includes insights and action logic. There is no exception in IIoT. The industry also places additional demands on security and interoperability.

The latest Gartner report forecasts 25.1 billion IoT units by the end of 2021. 80% of PoCs will not be converted into a full implementation with corresponding amortisation. The convergence of IT and OT (operational technology) and the increasing adoption of IoT in organisations have raised interoperability, integration and security issues that companies cannot avoid.

In the manufacture of machines, the various providers have agreed on an open standard - OPC Unified Architecture. OPC UA can be used regardless of platform and manufacturer. It improves communication from machine to machine and from machine to cloud. It is one of the building blocks of Industry 4.0.

Proof of concept (PoC) connection of customer machines

Before we get into the architectural details, I would like to remind you of the idea behind this PoC: It should provide us with important insights into some technical challenges that are expected for a future (SaaS) platform that is supposed to support millions of devices, thousands of users and a global reach. The PoC is intended to serve as one of the criteria for deciding whether and how the platform in question should be built.

In our PoC, we connected to the OPC UA server in two different ways. We connected two customer machines to the Swisscom IoT Cloud by using the Telit solution with our certified gateway device. We know that this solution is comprehensive. It can do many things, just to name a few benefits: It includes mobile connectivity with a machine-to-machine sim card. A broad interface for M2M services, including to AWS and Azure. Triggers for on-the-edge actions. High transaction reliability and so on.

The second way is via AWS IoT. We operate the device gateway with AWS Greengrass Core on the Raspberry Pi. An OPC UA client was developed and deployed as a long-running Lambda on the Greengrass Core. The data collection flow is the same as the Telit solution, but then the similarities end, because AWS IoT has a very different philosophy. It is a framework or toolsets. It doesn't offer us a plug-and-play solution. We have to put all the pieces of the puzzle together, sometimes even a bit of a puzzle. AWS IoT offers security and scalability by design, a high level of industrial acceptance and, last but not least, very active development of functions.

On both platforms, the data is forwarded to Kinesis Firehose. This is a data streaming service from AWS (a similar product to the open source project Kafka, for example). For the proof of concept, we simplify the process and download the data in S3 buckets. Once the data arrives, we perform some basic ETL processes (extract, transform, load) and write the output to the destination where our end customer can log in and see their machines and status. The green lines represent these processes (see graphic above). We have also developed "Update the thing by shadow". With this procedure, you can first update the device shadow (digital twin of the physical device) and perform some tests, e.g. change parameters such as "maximum allowable temperature". Then the new device configuration can be transferred from the device shadow to the real computer. We build the frontend with AWS Cognito for user management, D3.js and some Javascript to display the data interactively.

Lesson learnt

My personal experience is divided into more organisational/process-related topics on the one hand and technical topics on the other:

  • For successful IoT project implementation, an organisation must reach a certain level of maturity among all those involved. For example, IT and OT with different focal points must grow together into a joint responsibility model.
  • Managing different aspects is difficult. The IIoT touches on many different aspects such as data architecture, device management and edge aggregation. In addition, there is the software lifecycle, connectivity, manageability and security. Sufficient time should therefore be planned for these customisation tasks - close geographical collaboration could also be very helpful.
  • Take the issue of data ownership in the initial phase. The more data is collected, the more difficult it is to redistribute authorisations.
  • Build the architecture for the IIoT solution at an early stage. Use reference architectures from international bodies such as IIC or Industry 4.0 and not the architectures of individual providers.
  • Create and implement an end-to-end IoT view, and a test plan is the key to success!
  • By the way: If you want to build a completely new platform, as was the case in this project, choose a partner who will support you in developing the business concept before you decide on products and technologies for implementation. The business case should lead the discussions, not the technology.

If you have an IoT or IIoT project that you want to connect somehow to the global public clouds and need support, the Swisscom Global Public Cloud team can help you dive into AWS IoT and Azure IoT and more. Get in touch with our experts!(opens in new tab)

Wuming Zhang

Wuming Zhang

Cloud Architect

More getIT-articles

Ready  for  Swisscom

Find the job or career to suit you. A career where you can make a difference and continue your personal development.

What you do is who we are.

Go to careers

Go to current cyber security vacancies