Reference : Deep Learning to Predict the Feasibility of Priority-Based Ethernet Network Configurations
Reports : External report
Engineering, computing & technology : Computer science
Security, Reliability and Trust
http://hdl.handle.net/10993/44092
Deep Learning to Predict the Feasibility of Priority-Based Ethernet Network Configurations
English
Mai, Tieu Long mailto [University of Luxembourg > Faculty of Science, Technology and Communication (FSTC) > Computer Science and Communications Research Unit (CSC) >]
Navet, Nicolas mailto [University of Luxembourg > Faculty of Science, Technology and Communication (FSTC) > Computer Science and Communications Research Unit (CSC) >]
18-Aug-2020
University of Luxembourg
24
[en] Machine learning ; Graph Neural Network ; Design Space Exploration ; Time-Sensitive Networking ; Schedulability analysis ; In-vehicle networks
[en] This study is a contribution towards leveraging deep learning to further automate the design of communication architectures used in critical systems and, ultimately, design systems that are more efficient in terms of resource usage. Two well identified use-cases of deep-learning, and AI at large, in the design of critical systems are 1) fast prediction techniques that can replace, at some stages of the design, exact approaches, and 2) technology-agnostic configuration algorithms, {\it i.e.} algorithms not relying on extensive domain knowledge. This paper contributes to the first use-case and presents what is, to the best of our knowledge, the first deep learning model for feasibility prediction of real-time Ethernet networks.

Machine learning has been recently applied in real-time systems to predict whether Ethernet network configurations are feasible in terms of meeting deadline constraints without executing conventional schedulability analysis. However, the existing prediction techniques require domain expertise to choose the relevant input features and may perform poorly when topologies or traffic patterns differ significantly from the ones in the training data. To overcome these problems, we propose a Graph Neural Network (GNN) prediction model that synthesizes relevant features directly from the raw data. This deep learning model possesses the ability to exploit relations among flows, links, and queues in switched Ethernet networks, and, over the 13 testing sets used in this work, has proven an ability to generalize beyond the training data that is significantly superior to traditional ML algorithms.

We also explore the use of ensembles of GNNs and show that it enhances the robustness of the predictions. An evaluation on heterogeneous testing sets comprising actual automotive networks, shows that ensembles of 32 GNN models features a prediction accuracy ranging from 79.3% to 90% for Ethernet networks using priorities as the Quality-of-Service mechanism. The use of ensemble models provides a speedup factor ranging from 77 to 1715 compared to schedulability analysis. Such speed-up factors unlock new possibilities for design-space exploration and the development of near-interactive design tools. A practical advantage of our model is that it automates the feature engineering process, and does not require domain expertise. In that regard, the model could potentially be efficient in other areas of real-time computing.
Researchers ; Professionals ; Students ; Others
http://hdl.handle.net/10993/44092

File(s) associated to this reference

Fulltext file(s):

FileCommentaryVersionSizeAccess
Open access
TechReport_Graph_neural_networks_for_TSN.pdfAuthor preprint2.73 MBView/Open

Bookmark and Share SFX Query

All documents in ORBilu are protected by a user license.