IIC Journal of Innovation 10th Edition | Page 42

Keeping Ahead of the Curve with Custom ASICs Foundry Market by Feature Dimension 100 80 60 40 20 0 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 Year <10/7nm 16/14nm 20nm 28nm 45/40nm 65nm 90nm 0.13um 0.18/0.15um >0.25um Figure 3: Foundry Market by Feature Dimension decisions, the reality is that it is not always possible. As a result, there are excellent opportunities to leverage these technology nodes for the generation of custom ASICs for the IIoT and to do so at much lower costs than would have been thought previously. In real-time control, system processing can be performed at the source. This means that the data can be analyzed straight away, and decisions made quickly. A decision can then be made as to what data needs to be stored and what data must be sent to the cloud versus the data that is no longer required. This concept has become known as edge processing or edge computing. Edge processing involves bringing the processing power closer to the sensor edge where the physical world and the origins of digital data meet. Edge processing means response time is shortened and only the data that is truly required is transmitted. This helps to unburden the network and cut communication costs. E DGE C OMPUTING Earlier we discussed whether there is always a need to send all data to the cloud to achieve the desired business goals. In order to perform the level of monitoring needed, we need to install all the required sensors – and as already discussed, the numbers of sensors being deployed is growing at a large rate. In some cases, these sensors are placed in difficult to reach locations and running off battery power. Additionally, some remote locations mean communication constraints, so it is not always possible to communicate with the cloud when needed. So, while the ideal is to sense, measure and process all the data immediately and make intelligent IIC Journal of Innovation Edge processing is not without its difficulties, however. By bringing the processing tasks right to the sensor edge, there can quite often be space constraints. Also, as we - 38 -