A new model of network cascading failures with dependent nodes

Abstract

As real network systems are becoming ever more complex and widely-used, network failures often bring about devastating effects on our personal and professional lives. Therefore, failure regularity and reliability issues of real-world networks are primary concerns for network science researchers. In many real-world networks, cascading failures are not only caused by dynamical redistribution of network loads. Dependency relations among network nodes could also increase the spread of network failures, and even markedly impact the process of cascading failures. However, most of the existent cascading failure models used for network reliability analysis still merely focus on the dynamics of network loads for causing cascades without considering the joint effect of dependencies among network nodes. Based on previous studies and complex network theories, a new cascading failure model considering the synergies existing between dependency clusters of nodes and the dynamic distribution of network loads is proposed, upon which we analyze the effect of dependencies among network nodes on propagation of network failures. Firstly, we group network nodes into dependency clusters. Then, by triggering random break down of network nodes, we simulate cascading failures on random networks. The simulation results show that dependencies among network nodes will severely accelerate failure propagation and exacerbate the damage of cascading failures, in accordance with real situations. The obtained results firmly verify that the proposed cascading failure model describes the actual mixed failure behavior of networks with dependent nodes.