Fault tolerance designs of interconnection networks

Muhammad Faisal Nadeem, Muhammad Imran, Hafiz Muhammad Afzal Siddiqui, Muhammad Azeem

Research output: Contribution to journalArticlepeer-review

6 Citations (Scopus)

Abstract

Multiprocessor interconnection networks are mainly required to link a significant number of stable processor memory sets, every one of which is known as a processing vertex. Due to the cost-effectiveness, the design and utilization of multiprocessor interconnection networks have drawn attention lately. An interconnection network is extensively used in a parallel multiprocessor system to connect the processors and memory modules. In multiprocessor networks, vertices represent processors, and each processor of which is subject to complete failure. Fault tolerance is a process that ensures the working of a system to its total capacity, even if one or more of its components fail. In fault-tolerant design, backup components are used in interconnection networks to automatically take the place of the failed component, guaranteeing that the system’s working continues. This paper proposed the fault-tolerant design in the form of Pjk and Cjk graphs, where n processing components are linked, and i of these n components creating a chain of maximum size, are used to perform the task, they can tolerate the failure of components, maintaining steady operation. In particular, we present the fault-tolerant designs of pyramid, OTIS, biswapped, and mesh-derived networks.

Original languageEnglish
Pages (from-to)1125-1134
Number of pages10
JournalPeer-to-Peer Networking and Applications
Volume16
Issue number2
DOIs
Publication statusPublished - Mar 2023

Keywords

  • Biswapped network
  • Fault tolerance design
  • Interconnection networks
  • Mesh derived network
  • OTIS network
  • Pyramid network

ASJC Scopus subject areas

  • Software
  • Computer Networks and Communications

Fingerprint

Dive into the research topics of 'Fault tolerance designs of interconnection networks'. Together they form a unique fingerprint.

Cite this