Reverse Load Solicitation Methods for Fog Networks

This paper addresses the critical challenge of resource saturation in fog computing paradigms by proposing novel solicitation-based migration methods that alleviate near-saturated nodes. Unlike traditional reactive approaches that trigger migration post-saturation, the proposed methods employ continuous monitoring wherein adjacent nodes transmit beacon messages to advertise their instantaneous load levels. First, a dual least load and delay (LD) solicitation approach that balances resource utilization with link delay. Second, an affinity-based solicitation that migrates virtual network functions (VNFs) to nodes already hosting related elements from the same service function chain (SFC), thus preserving locality and reducing fragmentation. Simulations demonstrate that both methods outperform baseline proactive load balancing and reactive greedy search approaches at high traffic volumes in terms of saturation levels, migration iterations, service downtime, and SFC fragmentation. The affinity method particularly excels in minimizing SFC fragmentation and service disruption, while the LD approach provides saturation resistance across varying traffic volumes.