Digital twin models for predicting failure propagation in multi-vendor software ecosystems

Multi-vendor software ecosystems are essential components of critical digital infrastructure across finance, healthcare, manufacturing, cloud computing, and smart services. The rise of architectural modularity and integration using APIs, and third-party dependency chains has increased systemic vulnerability, where minor failures in single components may trigger broader service failures. Digital twin technology, which was initially developed based on cyber-physical systems, has become an exciting paradigm concerning the modeling of dynamic system behavior and its use in making a predictive decision [37], [38]. This paper presents an ecosystem-wide failure prediction framework, the Ecosystem Digital Twin of Failure Propagation (EDT-FP), which models distributed software ecosystems as time-dependent dependency graphs, augmented by telemetry-guided causal inferences as well as policy-conscious simulations. Experimental results indicated superior accommodation to propagation prediction, lower blast-radius estimation error, and improved mitigation ranking (as compared to topology-only or causal-only). The results suggest that an integrated digital twin architecture combining graph topology, multimodal telemetry, vulnerability modelling, and governance constraints can provide a scalable approach to achieve proactive resilience engineering in mixed and multi-vendor systems.