Availability evaluation of min-connected multiprocessors using decomposition technique

Chita R. Das, Lei Tien, Laxmi N. Bhuyan

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

An analytical technique for the availability evaluation of multiprocessors using a multistage interconnection network (MIN) is presented. The MIN represents a Butterfly-type connection with a 4 × 4-switching element (SE). The novelty of this approach is that the complexity of constructing a single-level exact Markov chain (MC) is not required. By use of structural decomposition, the system is divided into three subsystems--processors, memories, and MIN. Two simple MCs are solved by using a software package, called HARP, to find the probability of i working processing elements (PEs) and j working memory modules (MMs) at time t. A second level of decomposition is then used to find the approximate number of SEs (x) required for connecting the i PEs and j MMs. A third MC is then solved to find the probability that the MIN will provide the necessary communication. The model has been validated through simulation for up to a 256-node configuration, the maximum size available for a commercial MIN-connected multiprocessor.

Original languageEnglish (US)
Title of host publicationDigest of Papers - FTCS (Fault-Tolerant Computing Symposium)
PublisherPubl by IEEE
Pages176-183
Number of pages8
ISBN (Print)081862051X
StatePublished - 1990
Event20th International Symposium on Fault-Tolerant Computing - FTCS 20 - Chapel Hill, NC, USA
Duration: Jun 26 1990Jun 28 1990

Publication series

NameDigest of Papers - FTCS (Fault-Tolerant Computing Symposium)
ISSN (Print)0731-3071

Other

Other20th International Symposium on Fault-Tolerant Computing - FTCS 20
CityChapel Hill, NC, USA
Period6/26/906/28/90

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture

Fingerprint

Dive into the research topics of 'Availability evaluation of min-connected multiprocessors using decomposition technique'. Together they form a unique fingerprint.

Cite this