Green, reliable semiconductors are essential for Edge Computing. In conversation with Antonio Fernandez,Vice President Technical Development at EBV Elektronik.
Semiconductors are at the heart of each and every edge solution. However, the requirements for embedded systems differ considerably from those of computers or servers, as stressed by Antonio Fernandez, Vice President Technical Development at EBV. Yet according to him, semiconductor manufacturers are offering more and more solutions that have been specifically designed with edge computing in mind.
How do you define Edge Computing at EBV?
Antonio Fernandez: We understand Edge Computing as the infrastructure needed to implement high-performance data processing and storage close to the place where it is needed. At the edge of the network, near the data source. The application of artificial intelligence via the Internet of Things, the need of real-time performance with low response times, the need to preserve critical data privacy and reduce communication bandwidth for the massive amount of data produced by all the distributed sensors are key drivers in the demand for and evolution towards Edge Computing.
Will Edge Computing replace cloud computing or will both coexist?
A. F.: There will be a move from centralised IoT in the cloud to distributed IoT with Edge Computing; but both will coexist. Edge Computing will reduce the amount of data sent to the cloud and will reduce the latency of the network, improving response time. Edge solutions will enable safe, smart decisions in systems while they are not connected to the cloud and can improve data-privacy handling. The cloud will become much more sophisticated, managing massive volumes of data in real time. In the context of artificial intelligence, learning and big-data storage will happen in the cloud, while machine-learning sensing, inference and action will happen more frequently at the edge. Therefore, the edge will not replace the cloud; rather, it will complement it.
What is the role of embedded technology in that context?
A. F.: Embedded systems used in edge computing have special requirements that are different from personal computers or data-centre servers. They need to be compact, highly integrated, with several secure connectivity capabilities and interfaces to communicate with sensors and actuators. They have to be interoperable with other systems and need the right balance of computing performance and power consumption. The embedded technology must be robust and reliable enough to support specific environmental conditions and must have remote administration, monitoring and control. They have a convergence of computation, storage and network capabilities in common.
How important are semiconductors for embedded technology and especially for Edge Computing?
A. F.: The core of Edge Computing comprises the application processors, either as a stand-alone or integrated into FPGAs or application-specific devices. You also need memories for data handling and storage, interfaces, power management, timing, networking peripherals, etc. Using state-of-the-art, competitive, green and reliable semiconductors is essential for edge computing.
What are the main developments in semiconductor technology with regard to Edge Computing?
A. F.: The new generation of ARM application processors made with a 28-nanometre process offers higher performance, more peripheral integration, even including GPU, lower power consumption, higher reliability and very competitive prices. The pace of development, from FPGAs to systems-on-chip and adaptive compute acceleration platforms (so-called ACAPs) realised with cutting-edge 7-nanometre technology, offers capabilities that were previously out of reach. Their hardware and software are optimised for parallel heterogeneous computations and the integration of smart engines for AI, in addition to advanced digital signal processing.
In which areas is there a need for further development of hardware? Which trends do you see at your hardware suppliers?
A. F.: We are still missing the integration of machine-learning inference engines as a standard peripheral, but that is coming with a new generation of processors. Most of our hardware suppliers are getting ready for Edge Computing. The latest memories are matching the performance of the latest application processors, solid-state storage is meeting the cost target of the market, and peripherals are becoming highly optimised for latest-generation processors.
What is the role of software in this topic?
A. F.: Edge-Computing hardware without the proper software is useless. Software is the key enabler for getting the most out of the new peripherals, for exploding real-time processing, for implementing complex communication stacks, for realising security at the highest level and for ensuring interoperability between different systems. We even have software partitioning with cloud stacks, edge stacks and software stacks. They all need to work together with precision and reliability. Fortunately, there is an entire software ecosystem supporting the new design challenges, and a relevant part of the offering is free open software.
How can EBV support its customers in Edge Computing? What are your services and solutions?
A. F.: We are working very closely with our manufacturing partners to bring the latest technologies to the mass market in multiple applications and geographies. We have technical expertise in hardware and software that enables us to provide our customers with technical advice in product selection, live demonstrations, workshops and deep technical support to accelerate product development, solve technical problems and design for successful manufacturing.
What challenges are faced by manufacturers who want to make their products smart and equip them with Edge Computing capabilities?
A. F.: They will need to take care of security, interoperability between multiple connectivity and networking standards, real-time processing, remote management and reprogramming, and find a proper balance between computing performance, integration, cost and power consumption.
What role do open structures play in the further development of edge solutions? Where are the advantages?
A. F.: The latest open-hardware platforms and open-software ecosystems offer new levels of transparency, sustainability and cost reductions that benefit both the system manufacturer and the end user without compromising quality and reliability. It is a collective responsibility to enhance the tools so that everyone can benefit from the developments.
How will the market for edge solutions evolve? What are the big drivers?
A. F.: The market for edge solutions will grow very quickly as a consequence of the expansion of the Internet of Things empowered by Artificial Intelligence. We can imagine that, together with an increasing number of sensor and actuator nodes, there will be a large-scale deployment of smart gateways implementing Edge Computing due to several benefits. Firstly, it will reduce the volume of data going to the cloud and improve bandwidth. Secondly, selected data and insights flow to the cloud as required by data-protection laws. The third benefit is the possibility of near-real-time decisions and actions independent of cloud connections, thus, reducing unproductive waiting time. And, last but not least, intelligence incorporated into physical things and systems can give the user more added value.