Edge Computing: Where Calculations Really Help

Edge computing is a distributed IT architecture allowing data to be processed peripherally, as close to the source as possible. This makes up for latencies preventing the running times of applications from being close to real time

by Valerio Alessandroni

The increase in devices connected to the Internet (IoT) is producing a huge amount of data which need to be processed in data centres, pushing network bandwidth requirements to the limit. In spite of improvements in network technology, data centres cannot always guarantee acceptable transfer rates and response times. The problem is that the available bandwidth is not infinite and, notwithstanding the huge investments by providers in IT infrastructure, it risks not being adequate, especially for some specific requirements.

The advantages of edge computing with respect to cloud computing
Until recently, cloud computing was considered the traditional approach to meet IoT requirements. Cloud computing allows access to a shared set of computing resources (such as networks, servers, storage media, applications and services) with minimal interaction between the management centre and the service provider. However, the use of cloud computing as a centralised server, which is generally geographically distant, increases the frequency of communications between peripheral devices employed by users (tablets, computers, wristbands or smartphones) and becomes a limitation for applications requiring a real-time response. The process known as edge computing (or fog computing), which became very popular with the advent of Industry 4.0, is a distributed computing concept that brings computing and data storage closer to the position where it is required. This minimizes the need for long distance communications between clients and servers, improving latency (and therefore network performances) and allowing to save bandwidth.
Particularly, by processing data closer to the source and reducing the physical distance they need to cover, edge computing (or edge processing) optimizes Internet devices (IoT) and web applications. Basically, edge computing is a ‘network of micro data centres that process or store critical data locally in a very limited area’ (source: IDC).

The “network edge” for IoT and communication with the Internet
For IoT devices, the “network edge” is the point where the device, or the local network containing the device, communicates with the Internet. The edge is somewhat blurred: for instance, a user’s computer or a processor inside an IoT camera could be considered the edge of the network, but the user’s router, ISP, or local peripheral server could also be the edge. The important thing is keeping the edge of the network geographically close to the device, unlike traditional servers which can be very far away from the devices they communicate with. Let us consider a building protected by high-definition IoT cameras. These are normally “stupid” cameras that continuously transmit an unprocessed video signal to a cloud server. On the cloud server, the video signal from all cameras is filtered by a motion detection application to ensure that only the parts containing activity are stored in the server’s database. This means that there is constant and significant activity on the building’s Internet infrastructure because the high volume of video footage transferred uses up considerable bandwidth. In addition, there is a very heavy load on the server which needs to process video footage from all cameras simultaneously.
Let us now imagine moving the calculations performed by the motion sensors to the edge of the network. If each camera used its own internal processor to run the motion detection application and send the clips to the server as needed, bandwidth usage would be significantly reduced because much of the footage would no longer be transmitted. In addition, the server would only need to store important clips, so it could communicate with a larger number of cameras without risking an overload.

Privacy and security – what changes and what are the possible risks
Regarding privacy and security issues, the distributed nature of edge computing changes the schemes traditionally used in cloud computing: not only should data be encrypted, but different encryption mechanisms should be used, as they can pass through different distributed nodes connected via the Internet. On the other hand, by keeping the data on the edge, it is possible to move the property of collected data from service providers to end users.
Edge computing may however also have some disadvantages, such as the increase in potential security attacks. With the addition of more intelligent devices on the network, such as IoT devices with integrated processors, there are new opportunities for intruders. Another disadvantage is the demand for more local hardware.
For example, while an IoT camera needs an integrated processor to send unprocessed video data to a server, in order to run its own motion detection algorithms it would require a much more sophisticated computer with greater processing power.

A concrete initiative in Catania for Mobile Edge Computing
In December the first field test campaign of the European innovation project “Scenes – Smart City on the Edge Network Enhancements” took place in Catania.
The project aims to create an innovative IoT platform for Mobile Edge Computing, whereby a set of Intelligent Gateway devices (IGW) are positioned in vehicles travelling through the city; as they pass by, they interact with different sensors distributed all over the place, both to collect data to be sent to the service centre (the “Service Platform of Scenes”) and to remotely configure the sensors themselves. Another important objective is to create Scenes as an Open platform, where third parties, typically Smart City Service providers, can install their own applications on the Gateway, in order to bring the processing directly to the periphery, near the sensors deployed according to an edge computing approach.
Enabling the use of low cost sensors is another important project objective, which is believed to be an enabling factor for the creation of an ecosystem of new services and start-ups.
During the pilot test in Catania, version 1.0 of the Scene platform used for the monitoring of infrastructure and buildings using motion and vibration sensors was tested.
Six different sensors were positioned on some internal and external walls of a historical building: four commercial Bluetooth beacon accelerometers and two sensors equipped with accelerometer, gyroscope and thermometer.
The latter are low-cost wireless prototypes based on Raspberry technology, and allow to collect multiple performance parameters of the Scene platform. The data generated by all the sensors were sent to the IGW units installed inside three AMT (Azienda Metropolitana Trasporti) buses that periodically drove by the monitored building.
Within one of these buses the project team also carried out a series of tests on the Internet and Content Delivery Management features offered by the IGW unit installed on board.
During the tests hundreds of thousands of events were recorded by the platform and are now being brought to the attention of the Scene project researchers to be analysed.