IEEE International Conference on Computer Communications
We are delighted to inform you that Department of Computing PhD student Shreshth Tuli and his supervisor Dr. Giuliano Casale have received the best paper award at IEEE INFOCOM 2022, a premier international conference on state-of-the-art research in distributed systems and networks receiving over 1000 submissions every year.Their joint work, "PreGAN: Preemptive Migration Prediction Network for Proactive Fault-Tolerant Edge Computing", done in collaboration with Prof. Nicholas R. Jennings at Loughborough University, presents a novel model to ensure task resilience in edge computing environments. The PreGAN method leverages an adapted version of a Generative Adversarial Network (GAN) to perturb an incoming task placement decision and enact preemptive task migrations between edge devices, a technique useful to avoid resource exhaustion. Compared to the state-of-the-art methods, PreGAN reduces service level objective (SLO) violations by up to 12%, while incurring 24% lower execution overheads.PreGAN provides a step towards integrating data-driven AI based solutions in automated performance management of distributed systems. DoC's Quality of Service Research lab (QORE, https://qore.doc.ic.ac.uk/) is now extending PreGAN to ensure robust computing in situations where constraints exist on the data available to train the AI methods.
Article text (excluding photos or graphics) © Imperial College London.
Photos and graphics subject to third party copyright used with permission or © Imperial College London.
Reporter
Mr Ahmed Idle
Department of Computing
