Paper 07: Percolation Theory — Why Adding Redundancy Opens Highways for Cascading Outages

Tue, 22 Sep 2026 00:00:00 +0000

I. Abstract #

Intuition says: every replica added is an extra layer of safety, and these layers compound. Real systems do not behave that way. Every redundant link is a new edge on the graph. Any edge that routes service in one direction can route cascading failures in the other. Percolation theory reveals where the tipping point sits; load-dependent cascade models explain why failures propagate along those edges. A network dense enough to maintain service continuity is also dense enough to conduct outages across the entire cluster. Architectural resilience is therefore about choosing which edges to sever, not how many to add. Three measurable consequences: size headroom for post-failover load rather than peacetime load, catalog the shared dependencies that all replicas rely on, and prune the graph with cells rather than trusting replica counts.

Percolation Theory on VinhMDev

Paper 07: Percolation Theory — Why Adding Redundancy Opens Highways for Cascading Outages

I. Abstract #