Michael has more than 18 years of experience in different IT disciplines. He participated in the implementation of Hong Kong’s smart ID card system and Asia’s first RFID logistic system. With these experiences, Michael has authoritative insights in IoT and expertise in technology consulting.
Transformation Series - Solve Your Challenges on Data Analytics
Transformation Series - Solve Your Challenges on Data Analytics
Rethinking IoT architecture with Edge Computing
It’s 2019 and internet of things (IoT) is making headway in Hong Kong’s business landscape. More enterprises across industries are turning towards IoT for their digital initiatives. Shopping malls are using IoT to provide location-based services; property management companies are taking advantage of IoT to dynamically control lighting and air-conditioning for smarter and greener operations. All these services are building the foundation for Hong Kong to realise its smart city ambitions.
With sensor technologies maturing, adopting IoT has never been easier. However, the major challenge for successful IoT implementation is the management of sensors and connectivity with enterprise systems. Ensuring the sensors talk efficiently with the existing enterprise systems is key to bring analysis and meaning to data. Thus, IoT architecture needs to be able to handle large volumes of data traffic, real-time analysis and respond instantly to the data collected from different sensors.
The traditional IoT architecture powered by a centralised data centres are good at storing and processing data, but it lacks the capability to handle the surge in data traffic from hundreds or thousands of sensors. Meanwhile, technology advancement is creating powerful IoT devices, which can not only collect data but also process and analyse data. One example is facial recognition cameras, where computation power within the smart cameras can process some of the data collected at the edge of the IoT architecture.
These powerful IoT devices are driving the rise of edge computing in IoT. When data is processed and analysed closer to the source, the time and traffic required to transmit raw data to the core can be minimised. An IoT architecture with edge computing is particularly useful for services with zero tolerance to network delay or server downtime, like self-driving cars and airport or railway operation services. In the case of system outage at the core, edge computing could act as a short-term backup to process and store data.
More businesses that need to process data real-time are recognising the value of edge computing but often get lost in the IoT matrix. How much of the data processing and storage should be done at the edge? Should edge computing activities be placed at the sensor; or should there be fog computing - a separate edge computing device to process the data? These are some of the common questions. There are also concerns with regard to the time and resources required to build an IoT architecture that accommodates edge computing.
Between devices and the core, there is a wide spectrum of where and how edge computing can be placed. The answer depends on the requirements in performance and cost efficiency. For businesses with zero tolerance to delay, it is better to process data at the far edge at the sensors. The cost could be higher because more expensive and advanced smart devices are needed but real-time performance can be achieved.
It might not always be the most cost-efficient architecture to bring computing power to all the smart sensors. For instance, a smart building management system might only need one edge computing device on each floor to manage various sensors and controllers of the facility, like lighting, air conditioning, noise level, etc. The edge computing device can provide a fog computing architecture and act as an IoT middleware to manage and interoperate different types of sensors. This IoT architecture uses relatively cheaper sensors but requires more network and infrastructure support.
While some initial costs are inevitable when introducing a new system with edge computing capability, when smartly designed, it could help save considerable IT infrastructure costs in the long run.
Despite the popularity of cloud computing and the scalability it offers, not many are aware of the associated cloud subscription fee to process data from individual sensors. Public cloud service providers may require separate license fee for your fleet of IoT sensors. If there are 100 sensors, the license fee will be multiplied! By installing edge devices that collect, centralise and process data from IoT devices, it is easier to keep cloud subscription costs at a reasonable level.
On top of cloud subscription fees, connectivity is another major infrastructure cost item. The surge in data generated from sensors, especially those involving high volume of video data (e.g. facial recognition cameras), demands high bandwidth. Maturity of 5G networks may help to lower the connectivity cost, but the additional network bandwidth and data traffic cost could still be overwhelming. Installing edge devices could prevent the shock from receiving an overwhelming telco bill.
Like many emerging technologies, edge computing is still struggling with evolving standards. The lack of matured standards mean cookie-cutter solutions are yet to be available in the market. Thus, the role of solution architects is crucial. With the help of a solution architect, assessments can be made to help businesses design a customised IoT architecture that best fits their needs, eventually creating a perfect balance between performance and cost considerations.
Similar to ordering a tailor-made suit, designing a custom-made IoT architecture requires a thorough understanding of the customer’s needs. A suit design varies depending on its purpose, whether it is designed for a wedding, funeral or business. It is the same for customising an IoT architecture, the design varies depending on the purpose of the IoT system—whether it is part of a real-time facial recognition system or to monitor warehouse facilities—and the performance requirement.
Enterprises with a successful IoT system often involve business users in the design process. With better understanding of the business purposes and requirements, the IoT architecture design can incorporate business priorities, operations, and constraints.
Edge computing might not be the most prominent and exciting part of the IoT architecture, but it is an essential consideration to develop a cost efficient, productive and scalable IoT system. Businesses the world over are exploring edge computing as part of their IoT architecture to ensure success in connecting sensors, data, and core systems.