Scholarlink

<p>Fault diagnosis techniques for interconnection networks have gained significant attention in critical domains, such as data centers, parallel distributed computing, and cloud computing. However, existing diagnostic approaches predominantly address the single fault pattern, limiting their effectiveness in practical network environments characterized by complex fault distribution. To address this limitation, this article presents a dual-attribute fault diagnosis framework by systematically integrating the quantitative characteristics of network components with their topological scale attributes. First, the formalization of <Emphasis Type="BoldItalic">h</Emphasis>-extra <Emphasis Type="BoldItalic">r</Emphasis>-component diagnosability through rigorously theoretical analysis of fault distribution patterns applies to half-hypercube-based multiprocessor systems. Then, a novel Influence-Based Diagnosis Algorithm is designed under the Preparata-Metze-Chien (PMC) model, implementing dynamic fault detection analysis. Finally, experimental validation on both half-hypercube architectures and wiki-Vote networks demonstrates its superior performance. The proposed framework IBDA achieves 94.5% fault identification accuracy under large-scale testing scenarios, surpassing conventional single-attribute methods in terms of ACCR, TPR, TNR, FPR, FNR, Precision, F1_Score, MCC, and G-mean.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A dual-attribute fault diagnosis algorithm and its applications

  • 期刊论文

  • 发表日期：

    2025年8月28日

查看全文
  • Jiafei Liu,
  • Qi Wang,
  • Chia-Wei Lee

摘要

Fault diagnosis techniques for interconnection networks have gained significant attention in critical domains, such as data centers, parallel distributed computing, and cloud computing. However, existing diagnostic approaches predominantly address the single fault pattern, limiting their effectiveness in practical network environments characterized by complex fault distribution. To address this limitation, this article presents a dual-attribute fault diagnosis framework by systematically integrating the quantitative characteristics of network components with their topological scale attributes. First, the formalization of h-extra r-component diagnosability through rigorously theoretical analysis of fault distribution patterns applies to half-hypercube-based multiprocessor systems. Then, a novel Influence-Based Diagnosis Algorithm is designed under the Preparata-Metze-Chien (PMC) model, implementing dynamic fault detection analysis. Finally, experimental validation on both half-hypercube architectures and wiki-Vote networks demonstrates its superior performance. The proposed framework IBDA achieves 94.5% fault identification accuracy under large-scale testing scenarios, surpassing conventional single-attribute methods in terms of ACCR, TPR, TNR, FPR, FNR, Precision, F1_Score, MCC, and G-mean.