تحليل الأداء والوثوقية في شبكات الربط التفرعية المكعبية

Authors

  • روزا درويش قسم هندسة تكنولوجيا النظم الحاسوبية والإلكترونية، كلية هندسة تكنولوجيا المعلومات والاتصالات، جامعة طرطوس

Keywords:

Parallel Networks, Hypercube Topology, Fault Tolerance, Supercomputers, Interconnection Networks, Probabilistic Analysis, Redundancy.

Abstract

Hypercube interconnection networks are critical architectures in implementing high-integration circuits for high-performance parallel computing applications. This architecture features exceptional scalability and structural regularity, where each node possesses k input/output ports in a k-dimensional hypercube network. The main challenge lies in achieving optimal balance between fault tolerance and high performance, especially with increasing network sizes where Google studies have shown that clusters of 8,000 nodes suffer from failure rates reaching two-thirds of nodes annually (Google Inc., 2024).

This research aims to develop and analyze an enhanced hypercube network architecture (relative to the basic k-dimensional hypercube) using redundancy in both nodes and links, with complete software implementation using MATLAB. The results demonstrated a tangible improvement in reliability that escalates significantly with increasing failure probability — modest at low failure levels and substantial at high failure levels, particularly in higher-dimensional networks. This improvement is achieved against a constant and predictable increase in overall cost, while maintaining performance efficiency. Comparisons with other architectures proved the superiority of the enhanced hypercube network in critical applications requiring high reliability.

Downloads

Published

2026-10-29