Optimal restoration for resilience of dependency systems against cascading failures

Due to the increasing significance of complex and large-scale network systems and the potential for massive cascading failures in these systems, system resiliency modeling and optimal methods to restore systems from disruptions and to avoid further system loss are of significant interest among researchers and practitioners. Although past incidents have demonstrated the importance of cascading impacts aggravated by dependencies inside network systems, existing research on system resilience rarely if ever considers system dependencies and optimal restoration from cascading failures. In this paper, we applied different restoration strategies with respect to the strength of repairing actions to network systems, which suffered from mixed cascading failures that combines load dynamics and system dependencies. Based on multiple simulated networks, the effects of the presented restoration strategies were evaluated by system load fluctuation which tracks system evolution during the cascading process and two metrics of system resilience. Moreover, the influence of system dependencies with different dependency strength on system resilience was analyzed using numerical simulations. Our work improves understanding of practical restoration in response to cascading failures in order to enhance resilience of real-world network systems.