Publikation

An Introduction to Edge Computing and A Real-Time Capable Server Architecture

Volkan Gezer, Jumyung Um, Martin Ruskowski

In: International Journal on Advances in Intelligent Systems (IARIA) 11 1&2 Seiten 105-114 IARIA 7/2018.

Abstrakt

The Internet has changed the way people access the information they need, and indeed how they live. Whether it is individuals reading emails or watching videos, or factories utilising automated fabrication devices, the access and processing of data is totally different. Thanks to the accessibility and the benefits that it brings into the lives, new research areas are emerging. One of the areas is Internet of Things (IoT) which connects countless of devices to the Internet. Increasing usage in IoT tremendously increases the count of connected devices to the Internet as well as the data generated and transferred. However, this increase brings several issues which could degrade the Quality of Service (QoS) with delays or even failed requests due to bandwidth limitations. Current tendency to solve problems that the Cloud Computing has is to perform computations close to the device as much as possible. This paradigm is called Edge Computing. There are several proposed architectures for the Edge Computing, but there is not an accepted standard by the community or the industry. Besides, there is not a common agreement on how Edge Computing architecture physically looks like. In this paper, we describe the Edge Computing, explain how its architecture seems, its requirements, and enablers. We also define an extensible, server architecture. The proposed Edge Server architecture has an ability to decide whether the task should be offloaded to the Cloud or to another Edge Server by considering the several parameters such as available resources and network delays. The resources of Edge Server can be extended with additional optional hardware or software modules to add new functionalities for artificial intelligence tasks, additional storage, wireless communication, etc. The server in the proposed architecture is also capable of performing real-time tasks and uses standard technologies to keep migration efforts at minimum. The paper also shows the results of an initial experiment, done without and with an Edge Server to compare computing performance.

Weitere Links

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence