Passive bandwidth estimation for fog computing

Abstract

This paper investigates passive bandwidth estimation as a
means to improve scheduling in Kubernetes-based fog computing environments.
In contrast to active probing techniques, passive methods
rely on observing existing traffic patterns to infer available bandwidth
(ABW), offering a non-intrusive alternative suitable for dynamic
and distributed systems. The study focuses on wired infrastructure networks
within the cloud–fog–edge continuum.

We develop and present an approach based on a modified version of
the probe gap model, evaluated in an emulated network testbed. Ten
experiments were conducted under varying traffic conditions, packet loss
rates, and link capacities. Timestamping artifacts and data filtering were
addressed to ensure measurement integrity. The approach demonstrates
reliable ABW estimation across all scenarios, confirming the viability of
passive methods for bandwidth-aware scheduling in Kubernetes.