Resumo

VALDES DIAZ, Beatriz  e  ARTIGAS FUENTES, Fernando José. C++ library for work with clusters. Rev cuba cienc informat [online]. 2020, vol.14, n.2, pp.15-33.  Epub 01-Jun-2020. ISSN 2227-1899.

Computer clusters are important for the development of complex scientific research. Some studies demonstrate that tools used on clusters are complicated for users less related to technical sciences. This paper presents the design and implementation of a C++ library targeted to programmers, which arranges the work with a cluster into layer architecture. This allows programmers to use high-level abstraction functions, closer to the user's problem to be solved. The library allows not only integrate user´s desktop applications with clusters but also makes part of the work with them in an unattended way and speeds up scalability studies of parallel programs. Likewise, specifies cluster resources and job features required to run the programs by using predefined recipes. As part of the implementation, the job manager component of the GridMD library, and the boost_serialization and boost_filesystem libraries were used. All the tests carried out to the library on a cluster with Linux and Slurm have shown a good performance without identified errors.

Palavras-chave : library; cluster; layered architecture; HPC; C++.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )