A root-cause localization framework known as Decoupled Multiplex Causal Ranking has demonstrated improved accuracy in identifying failure origins across microservice systems. Failures in these distributed architectures produce correlated symptoms across services, motivating the use of structural information alongside anomaly-based approaches. In benchmark evaluations, the framework’s core algorithm, Regime-aware Innovation Causal Estimation with Confidence-aware Multiplex Coupling (RICE-CM), achieved the highest average top-five ranking accuracy among all evaluated methods across three benchmark datasets. On the Online Boutique benchmark, the algorithm exceeded the competing RUN method by an absolute margin of 0.0288 while requiring approximately one thirty-third of RUN’s runtime per case.
The study investigated service-level localization from multivariate monitoring metrics under explicit operational assumptions, specifically known fault-onset timing and known metric-to-service mappings. Despite this need, existing temporal and multilayer ranking methods rarely use relation-estimation reliability to control node-level information exchange between contemporaneous and lagged evidence.
To address these limitations, the authors proposed Decoupled Multiplex Causal Ranking alongside its core algorithm, Regime-aware Innovation Causal Estimation with Confidence-aware Multiplex Coupling. The algorithm derives both contemporaneous and lagged relational evidence from a shared dynamics model fitted to the pre-fault normal operational segment.
Methodologically, RICE-CM aggregates edge confidence scores into node reliability values and applies nonuniform diagonal coupling between same-metric replicas across multiplex layers. By utilizing node reliability to govern information exchange between contemporaneous and lagged evidence, the method avoids fixed inter-layer coupling assumptions. Service root causes are subsequently ranked by performing reverse PageRank on the assembled supra-adjacency matrix.
The researchers evaluated RICE-CM on a 300-case subset drawn from three benchmark datasets in RCAEval: Sock Shop 1 (SS1), Sock Shop 2 (SS2), and Online Boutique (OB). With dataset-specific coupling settings selected on these benchmark cases, RICE-CM achieved Avg@5 scores of 0.6000 on Sock Shop 1, 0.7040 on Sock Shop 2, and 0.4864 on Online Boutique. These scores represented the highest Avg@5 ranking accuracy among all evaluated methods across all three benchmark datasets.
Compared with the ablation baseline RICE-only, RICE-CM improved Avg@5 accuracy by 0.0640 on Sock Shop 1, 0.0704 on Sock Shop 2, and 0.0624 on Online Boutique. Paired statistical tests supported these measured gains on Sock Shop 1 and Sock Shop 2 after Holm correction, although the improvement on Online Boutique did not reach statistical support after the same correction. Nevertheless, positive observed gains over RICE-only occurred across all twelve dataset-fault-type combinations evaluated in the experiments.
Against the fixed-coupling baseline RICE-MX, RICE-CM produced Avg@5 improvements of 0.0720 on Sock Shop 1, 0.0400 on Sock Shop 2, and 0.0336 on Online Boutique, with a Holm-adjusted p-value of 0.0456 on each dataset. However, coupling-parameter sweeps revealed dataset-dependent sensitivity across the benchmark tests. These empirical findings substantiate the efficacy of node-reliability-driven nonuniform coupling over fixed coupling specifically within the evaluated benchmarks and operational task assumptions.
Subject of Research: Microservice root cause analysis using decoupled multiplex causal ranking
Article Title: DeCoR: decoupled multiplex causal ranking for microservice root cause analysis
Article References: Zheng, Y., Zhao, M., Wan, L., & Yao, H. (2026). DeCoR: decoupled multiplex causal ranking for microservice root cause analysis. Journal of Cloud Computing. https://doi.org/10.1186/s13677-026-01009-x
Image Credits: AI Generated
DOI: 10.1186/s13677-026-01009-x
Keywords: microservice systems, root cause analysis, causal ranking, DeCoR, RICE-CM, multiplex coupling, multivariate monitoring metrics, reverse PageRank, supra-adjacency matrix, RCAEval, Sock Shop, Online Boutique, Avg@5, RUN baseline, RICE-MX, node reliability, Holm correction
News Source: Denise Maddox. (November 18, 2026). Decoupled Multiplex Causal Ranking Improves Microservice Root Cause Localization in Benchmarks. Scienmag.



