WARNING: This is the _old_ Lustre wiki, and it is in the process of being retired. The information found here is all likely to be out of date. Please search the new wiki for more up to date information.

Lustre Publications: Difference between revisions

From Obsolete Lustre Wiki
Jump to navigationJump to search
Line 26: Line 26:
* http://www.rz.uni-karlsruhe.de/dienste/lustretalks.php
* http://www.rz.uni-karlsruhe.de/dienste/lustretalks.php
* Six talks in PDF
* Six talks in PDF
** Einführungsveranstaltung im Rechenzentrum (2005):
** Einführungsveranstaltung im Rechenzentrum (2005): [http://wiki.lustre.org/images/7/7c/Karlsruhe0503.pdf Karlsruhe0503.pdf]
** HP-CAST 4 in Krakau (10.5.2005):
** HP-CAST 4 in Krakau (10.5.2005): [http://wiki.lustre.org/images/9/95/Karlsruhe0510.pdf Karlsruhe0510.pdf]
** ISC 2005 in Heidelberg (24.6.2005):
** ISC 2005 in Heidelberg (24.6.2005): [http://wiki.lustre.org/images/1/17/Karlsruhe0506.pdf Karlsruhe0506.pdf]
** HP-CAST 5 in Seattle (11.11.2005):  
** HP-CAST 5 in Seattle (11.11.2005): [http://wiki.lustre.org/images/1/17/Karlsruhe0511.pdf Karlsruhe0511.pdf]
** HP-CCN in Seattle (12.11.2005):  
** HP-CCN in Seattle (12.11.2005): [http://wiki.lustre.org/images/1/17/Karlsruhe0512.pdf Karlsruhe0512.pdf]
** SGPFS 5 in Stuttgart (4.4.2006):
** SGPFS 5 in Stuttgart (4.4.2006): [http://wiki.lustre.org/images/1/17/Karlsruhe0604.pdf Karlsruhe0604.pdf]


== Ohio State University ==  
== Ohio State University ==  

Revision as of 04:36, 22 May 2007

Caspur (CERN)

  • attachment
  • we are doing poorly on 64k random reads, panasas does much better.

Cray User Group

  • XT7? Integrating and Operating a Conjoined XT3+XT4 System
    • Presentation: Presented by ORNL on CUG 2007
    • Paper:This paper describes the processes and tools used to move production work from the pre-existing XT3 to the new system incorporating that same XT3, including novel application of Lustre routing capabilities.
    • docs on CUG site Paper Presentation

HEPiX Talks

Karlsruhe Lustre Talks

Ohio State University

  • Lustre performance comparison when using InfiniBand and Quadrics interconnects
  • Download paper at OSU site: [2]
  • attachment:

University of Boulder, Colorado

  • LCI paper: attachment:
    • Note that when a single client gets exclusive access GPFS wins
    • Single client reads (we are very poor): remedy - tune
    • Creates in a unique directory per client: remedy - memory WB cache
    • The management issues mentioned in the last part of this paper are being addressed
    • [3](It's the same as the attachment of LCI paper above.)

University of Minnesota

  • attachment:MSST-2006-paper.pdf