Paper accepted at EdgeSys'23

12 April 2023, by Heiko Bornholdt

Photo: https://edge-sys.github.io/2023

Our paper, ”Latency-Aware Scheduling for Real-Time Application Support in Edge Computing” was accepted for publication at the 6th International Workshop on Edge Systems, Analytics and Networking (EdgeSys 2023), co-located with EuroSys ‘23.

In this paper, we provide a comprehensive analysis of edge offloading and its associated latency behavior. For that, we conducted real-world experiments, whose results we describe in detail.

We will present our results at the workshop in Rome in May 2023 and seek an exchange with international researchers.

Paper Abstract:

The relocation of computation from the network core to the edge where data is primarily generated has gained momentum, leading to the emergence of edge computing as a viable solution for low-latency processing. As a result, edge computing has the potential to significantly reduce response times, decrease bandwidth usage, enhance energy efficiency, and offer various other benefits. At the same time, end-user devices do not offer a consistent computing platform and Internet middleboxes severely restrict communication with edge devices. Often, this is circumvented by publicly accessible relay servers, which cause additional latency and render time-critical tasks unviable for offloading. This paper presents an approach capable of addressing the complexities inherent in edge computing and facilitating good decisions regarding latency-aware computation offloading. We conducted several real-world experiments to evaluate our approach and provide valuable data for further research. Our findings show that edge offloading is competitive to cloud and grid offloading, as it effectively reduces latency. Empirical evidence from our research supports that edge computing can offer significant advantages for real-time applications.