Edge Offloading for Microservice Architectures

Edge Offloading for Microservice Architectures

Abstract

Edge offloading is widely used to support the execution of near real-time mobile applications. However, offloading on edge infrastructures can suffer from failures due to the absence of supporting systems and environmental factors. We propose a fault-tolerant offloading method modeled as a Markov Decision Process (MDP) based on predictions performed through Support Vector Regression (SVR). SVR is used to estimate offloading service availability, which is used by MDP for offloading decisions. Our approach is implemented in a real-world test-bed and compared with the default Kubernetes scheduler augmented with hybrid fault-tolerance.

Grafik Top
Authors
  • Zilic, Josip
  • De Maio, Vincenzo
  • Aral, Atakan
  • Brandić, Ivona
Grafik Top
Shortfacts
Category
Paper in Conference Proceedings or in Workshop Proceedings (Paper)
Event Title
5th International Workshop on Edge Systems, Analytics and Networking (EdgeSys 2022)
Divisions
Scientific Computing
Subjects
Software Engineering
Kuenstliche Intelligenz
Systemarchitektur Allgemeines
Event Location
Rennes, France
Event Type
Workshop
Event Dates
05.04.2022
Date
5 April 2022
Export
Grafik Top