A database for on-line event analysis on a distributed memory machine

Parallel in-memory databases can enhance the structuring and parallelization of programs used in High Energy Physics (HEP). Efficient database access routines are used as communication primitives which hide the communication topology in contrast to the more explicit communications like PVM or MPI. A...

Descripción completa

Detalles Bibliográficos
Autores principales: Argante, E, Meesters, M R J, Van der Stok, P D V, Willers, Ian Malcolm
Lenguaje:eng
Publicado: 1995
Materias:
Detectors and Experimental Techniques
Acceso en línea:http://cds.cern.ch/record/294828
Descripción
Sumario:Parallel in-memory databases can enhance the structuring and parallelization of programs used in High Energy Physics (HEP). Efficient database access routines are used as communication primitives which hide the communication topology in contrast to the more explicit communications like PVM or MPI. A parallel in-memory database, called SPIDER, has been implemented on a 32 node Meiko CS-2 distributed memory machine. The spider primitives generate a lower overhead than the one generated by PVM or PMI. The event reconstruction program, CPREAD of the CPLEAR experiment, has been used as a test case. Performance measurerate generated by CPLEAR.

Ejemplares similares