Please note that conversion to PDF is currently not available. More details here
2-6 May 2011
GSI Helmholtzzentrum für Schwerionenforschung GmbH
Europe/Berlin timezone

Evaluation of distributed file systems using trace and replay mechanism

4 May 2011, 12:15
30m
Hörsaal / lecture hall (GSI Helmholtzzentrum für Schwerionenforschung GmbH)

Hörsaal / lecture hall

GSI Helmholtzzentrum für Schwerionenforschung GmbH

Planckstr. 1, 64291 Darmstadt, Germany
Storage & Filesystems Storage & FileSystems

Speaker

Mr Jiri Horky (Institute of Physics of Acad. of Sciences of the Czech Rep. (ASCR))

Description

Reliable benchmarking of file systems is a complex and time consuming task when one has to test against a production environment to achieve relevant results. In case of the HEP community, this eventually leads to setting up a particular experiments' software environment, which could be a rather complicated task for a system administrator. To simplify this task, we developed an application for exact replaying of IO requests to reliably replicate an IO behavior of the original applications without a need of installing the whole working environment. Using the application, we present performance comparison of Lustre, GPFS and Hadoop file systems by replaying traces of LHCb, CMS and ATLAS jobs.

Primary author

Mr Jiri Horky (Institute of Physics of Acad. of Sciences of the Czech Rep. (ASCR))

Presentation Materials