Hybrid MPI/OpenMP cloud parallelization of harmonic coupled finite strip method applied on reinforced concrete prismatic shell structure. (June 2015)

• Record Type: Journal Article
• Authors: Nikolić, M.; Hajduković, M.; Milašinović, D.D.; Goleš, D.; Marić, P.; Živanov, Ž.
• Is Part Of: Advances in engineering software. Volume 84(2015)
• Page Start: 55
• Page End: 67
• Publication Date: 2015-06
• Subjects: Cloud computing -- MPI -- OpenMP -- Harmonic coupled finite strip method -- Shell prismatic structures -- Reinforced concrete; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2014.12.006
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: Application of hybrid parallelization model coupling MPI with OpenMP. Rational usage of cloud computer resources in hybrid parallelization. Analysis of impact of cluster nodes and CPU cores number to parallelization effects. Harmonic coupled finite strip method application on shell structures. Method demonstrated on reinforced concrete prismatic shell structure. Abstract: This paper discusses the cloud computing based approach for parallelization of large displacement stability analysis of orthotropic prismatic shell structures with simply supported boundary conditions along the diaphragm-supported edges. We review the harmonic coupled finite strip method (HCFSM), and describe a software system for nonlinear analysis of reinforced concrete (RC) structures. We combine different parallelization models – MPI and OpenMP – in order to cope with the increased computational complexity, which originates from coupling of all series terms in the HCFSM formulation. We discuss the effects of parallelization from the perspective of a cloud environment. Our results show that rational usage of cloud resources can lead to significant performance improvements and monetary savings. In certain cases, the achieved performance can be very close to the maximum one.