Adeunis and CARL Software experts are working together to develop an IoT sensor integrating IA and Edge Computing.
Jean-Luc Baudouin Deputy Managing Director at Adeunis and Youssef Miloudi, Product Technical Manager at CARL Softawre, took part in the 3 questions on Edge Computing as part of this development.
What is Edge Computing?
Jean-Luc Baudouin : Edge Computing is both an approach and a set of information distribution, processing techniques and technologies that support the collection and analysis of data in real time, locally (in the sensor) or at the sensor periphery.
Why Edge Computing?
Youssef Miloudi : Edge Computing is the answer to the problem of optimizing the processing of massive data. Indeed, by processing the majority of data as close as possible to its original source, before transmitting it to the cloud via the network, Edge Computing will become essential for the scale-up, say on a large scale, of IoT solutions. Forecasts estimate that by 2020, there will be approximately 50 billion connected objects in the world. All these objects will generate a lot of data that needs to be transported, stored, pre-processed and analyzed.
In addition, for the many industrial cases where analysis must be done in near-real time with rapid decision making, the problems of latency and delayed decisions due to batch processing in the cloud are crippling.
Jean-Luc Baudouin : In summary, regarding IoT data processing there are two common approaches:
- Continuous massive collection / dataLake = all the captured data is sent, continuously or in “batch“, and the analysis of their relevance to a given use case will be done in a 2nd step, in the cloud.
- Data Sets: data is processed and transformed locally into relevant business information for decision making. The objective of local processing is to allow only relevant and useful data to be sent. Edge computing corresponds to this approach.
What are the benefits of Edge Computing?
Youssef Miloudi : Edge Computing, by considerably reducing the volumes of data that transit, makes IoT solutions more efficient and more economical. It reduces latency and associated costs, improves security and speeds up decision making.
Limitation of latency in local decision making and equipment control
Jean-Luc Baudouin : With Edge Computing, all data no longer must be sent to the cloud before being processed. The greater the distance between the places where data is created, collected, processed and analyzed, the longer the processing time. With Edge Computing, the path between data creation and processing is minimized, so the time required to process and analyze the data is also reduced. The information sent is also more relevant and therefore its processing is more efficient and requires less pre-processing (cleaning, aggregation, etc…).
For example, for equipment maintenance, while a classic IoT device will transmit measurements that need to be analyzed and qualified in the cloud (or a business application), a device with Edge Computing will immediately qualify the alarms or symptoms related to the measurements and transmit qualified information to the business application.
This “saved” time allows for faster decision making and improved operational performance.
Limitation of energy expenditure
Jean-Luc Baudouin : Today’s IoT applications generate large volumes of data, frequently sent to the cloud and much of this data volume is deleted or becomes useless after processing! The cloud is “polluted” by a lot of “useless or temporary” data and consumes a lot of energy to store this data. By processing data locally, only relevant data is transferred to the cloud. The costs of transmission, storage and processing are therefore greatly reduced, as well as the environmental impact.
Maintaining quality of service
Youssef Miloudi : With Edge Computing, radio frequency bands are not unnecessarily occupied by large amounts of data that could saturate and potentially lead to a decrease in the quality of service of IOT products.
Limitation of security risks
Jean-Luc Baudouin : By not sending all data over the networks but only useful and therefore valuable information with context elements known only to the users, the security risks are limited.