Conference Presentation (Other) FZJ-2018-00330

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png
The NESTio project – replacement data recording backend for NEST

 ;  ;

2017

NEST Conference 2017, JülichJülich, Germany, 19 Dec 2017 - 20 Dec 20172017-12-192017-12-20

Please use a persistent id in citations:

Abstract: Simulation data is recorded in NEST by virtual spike detectors and multimeters. In regular intervals during NEST simulations, the recorded data is written to the file system. The current implementation of this writing process does not scale well to large simulations which run on large clusters or supercomputers with several hundreds to tens of thousands of compute nodes. For this reason, a completely new software architecture for the recording backends of NEST was developed as the nestio project. This architecture allows the user to direct recording device output to various different file formats and interfaces. Particularly appropriate for large-scale simulations on supercomputers is the implementation of a backend writing to sionlib container files. Benchmarking results show that the sionlib recording backend reduces simulation times by a considerable factor compared to the current NEST recording backend whenever large amounts of data are written to the file system during simulations.


Contributing Institute(s):
  1. Jülich Supercomputing Center (JSC)
Research Program(s):
  1. 511 - Computational Science and Mathematical Methods (POF3-511) (POF3-511)
  2. SMHB - Supercomputing and Modelling for the Human Brain (HGF-SMHB-2013-2017) (HGF-SMHB-2013-2017)
  3. SLNS - SimLab Neuroscience (Helmholtz-SLNS) (Helmholtz-SLNS)

Appears in the scientific report 2017
Database coverage:
OpenAccess
Click to display QR Code for this record

The record appears in these collections:
Dokumenttypen > Präsentationen > Konferenzvorträge
Workflowsammlungen > Öffentliche Einträge
Institutssammlungen > JSC
Publikationsdatenbank
Open Access

 Datensatz erzeugt am 2018-01-11, letzte Änderung am 2021-01-29


Dieses Dokument bewerten:

Rate this document:
1
2
3
 
(Bisher nicht rezensiert)