Binomial graph: a scalable and fault-tolerant logical network topology

  • Authors:
  • Thara Angskun;George Bosilca;Jack Dongarra

  • Affiliations:
  • Dept. of Computer Science, The University of Tennessee, Knoxville;Dept. of Computer Science, The University of Tennessee, Knoxville;Dept. of Computer Science, The University of Tennessee, Knoxville

  • Venue:
  • ISPA'07 Proceedings of the 5th international conference on Parallel and Distributed Processing and Applications
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

The number of processors embedded in high performance computing platforms is growing daily to solve larger and more complex problems. The logical network topologies must also support the high degree of scalability in dynamic environments. This paper presents a scalable and fault tolerant topology called binomial graph (BMG). BMG provides desirable topological properties in terms of both scalability and fault-tolerance for high performance computing such as reasonable degree, regular graph, low diameter, symmetric graph, low cost factor, low message traffic density, optimal connectivity, low fault-diameter and strongly resilient. Several fault-tolerant routing algorithms are provided on BMG for various message types. More importantly, BMG is able to deliver broadcast messages from any node within log2(n) steps.