Abstract
Modern Continuous Integration and Continuous Deployment (CI/CD) systems rely predominantly on static analysis and deterministic rules, rendering them incapable of forecasting emergent risks arising from complex infrastructure interactions. This reactive approach frequently fails to prevent production incidents caused by non-obvious dependencies between code, configuration, and infrastructure. This paper proposes a novel Artificial Intelligence for IT Operations (AIOps) system that models cloud infrastructure as a topological dependency graph. We introduce the concept of a "Delta Graph" ($\Delta G$), which explicitly represents the structural and attribute changes between a currently deployed state ($t_0$) and a proposed state ($t_1$). By utilizing a Graph Neural Network (GNN) trained on historical delta graphs, the system performs dependency-aware risk propagation to compute a Rollback Probability Score (RPS). This dynamic risk assessment allows the CI/CD pipeline to proactively block high-risk, multi-faceted changes that would otherwise bypass conventional static checks, effectively preventing incidents before deployment.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
BHARDWAJ, UTKARSH and Awasthi, Shivank, "A Graph-Based Infrastructure Change Modeling System with Dependency-Aware Risk Propagation for Pre-Deployment Failure Prediction", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11784