Experimental Platform for Edge Computing Applications
- Diarienummer
- ITM17-0246
- Start- och slutdatum
- 190401-230331
- Beviljat belopp
- 7 933 777 kr
- Förvaltande organisation
- KTH - Royal Institute of Technology
- Forskningsområde
- Informations-, kommunikations- och systemteknik
Summary
While data networks are highly optimized towards today’s applications, novel closed-loop applications are emerging which require much improved network performance. Closed-loop applications refer to AI-supported human assistance systems or automatic control applications. To warrant their seamless integration into reality, most importantly the network responsiveness must be improved drastically. This can only be achieved by providing computation services as close as possible to the application, referred to as edge computing. While edge computing is an important emerging architecture, little is known about its implications for closed-loop applications. Essentially, there is a gap between theoretical work and the achievable performance of edge computing systems in practice. It is very important to overcome this gap, which is the core motivation for this project. Through establishing a highly reprogrammable wireless networking and edge computing infrastructure, as well as providing standardized benchmarks and use cases, we hope to provide an open-access experimental infrastructure that will allow the validation of theoretical results in practice. In addition, this testbed will facilitate an environment for deriving corresponding architecture and interface consequences, as well as seeding the formation of an ecosystem in the strategically important intersection between dependable applications and emerging edge computing infrastructures.
Populärvetenskaplig beskrivning
Over the last twenty years communication networks and their applications have had a tremendous impact on societies around the world. Today, these communication networks are mainly understood as transportation infrastructures for information, i.e. bit pipes, which work sufficiently well for various applications like voice and video communication as well as web services. In contrast, novel upcoming so-called closed-loop applications will turn data networks into compute infrastructures where artificial intelligence algorithms will interact with reality in an autonomous way. Examples of such applications are augmented reality or also control applications in industrial manufacturing. However, in order to realize this vision, the responsiveness of networked infrastructures needs to be improved drastically, leading to the concept of edge computing. In edge computing, computational elements are provided as close to the application end points as possible, perhaps at the base station of wireless networks. This leads to much reduced latencies and improves therefore the responsiveness. While this evolution is quite clear, much of its implications are not well understood. In theory, models can be established which clearly allow the determination of the level of responsiveness achievable in a certain environment. However, these models are well known to be quite abstract with only limited practical relevance. This motivates us to establish an experimental platform, which can be employed to investigate such divergences between theory and reality, with important lessons to be learned for future networked infrastructures over which closed-loop applications will be run. This project will realize an open-access experimental testbed, which comprises of highly reprogrammable communication and computation components, and which will furthermore rest on the definition of standardized and reproducible use cases. This platform will hence serve direct research addressing the gap between theory and practice, will pave the way towards practically relevant architectures, while also serving as a starting point for the formation of an academic and industrial ecosystem. Closed-loop applications have a high strategic relevance for the Swedish industry and we thus hope therefore to contribute to further strengthen the related stakeholders in academia and industry, and therefore boost the competitiveness of the Swedish society in the respective area.