I''m trying to modify lustre cache parameters, but I don''t find info. The problem is that if I open a 1''5 Gb file to read in ext3, the second time, system can read it in 0.9 seconds, but if I open it in lustre, second time takes 11.2 seconds. It seems that lustre doesn''t use right virtual file system? Do you know which parameters should I must change?, I don''t find under /proc valid parameters to change it ---------------------------------------------------------------------- Mensaje enviado por el servidor de openSistemas (www.opensistemas.com)
On May 29, 2006 17:26 +0200, jmgonzalez@opensistemas.com wrote:> I''m trying to modify lustre cache parameters, but I don''t find info. > The problem is that if I open a 1''5 Gb file to read in ext3, the second time, > system can read it in 0.9 seconds, but if I open it in lustre, second time > takes 11.2 seconds. It seems that lustre doesn''t use right virtual file system? > > Do you know which parameters should I must change?, I don''t find under /proc > valid parameters to change itHow large is RAM? There is /proc/fs/lustre/llite/fs0/max_cache_mb, which is default to 3/4 of RAM size. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.
It hasn''t RAM problem, it''s a pseries partition and has 8 Gb RAM. The idea is to use Lustre in a SAN and switch only two servers, so it''s so important cache. The servers have a lot of RAM, and we would like, they''d have almost all data in cache. Mensaje citado por Andreas Dilger <adilger@clusterfs.com>:> On May 29, 2006 17:26 +0200, jmgonzalez@opensistemas.com wrote: > > I''m trying to modify lustre cache parameters, but I don''t find info. > > The problem is that if I open a 1''5 Gb file to read in ext3, the second > time, > > system can read it in 0.9 seconds, but if I open it in lustre, second time > > > takes 11.2 seconds. It seems that lustre doesn''t use right virtual file > system? > > > > Do you know which parameters should I must change?, I don''t find under > /proc > > valid parameters to change it > > How large is RAM? There is /proc/fs/lustre/llite/fs0/max_cache_mb, which is > default to 3/4 of RAM size. > > Cheers, Andreas > -- > Andreas Dilger > Principal Software Engineer > Cluster File Systems, Inc. > >---------------------------------------------------------------------- Mensaje enviado por el servidor de openSistemas (www.opensistemas.com)
Hi, The Lustre OSS servers don''t cache data. We were curious how much RAM you had on the Lustre client which is reading the file. - Peter -> -----Original Message----- > From: lustre-discuss-bounces@clusterfs.com > [mailto:lustre-discuss-bounces@clusterfs.com] On Behalf Of > jmgonzalez@opensistemas.com > Sent: Tuesday, May 30, 2006 2:19 AM > To: lustre-discuss@clusterfs.com > Subject: Re: [Lustre-discuss] lustre performance > > It hasn''t RAM problem, it''s a pseries partition and has 8 Gb RAM. > The idea is to use Lustre in a SAN and switch only two > servers, so it''s so important cache. The servers have a lot > of RAM, and we would like, they''d have almost all data in cache. > > > > Mensaje citado por Andreas Dilger <adilger@clusterfs.com>: > > > On May 29, 2006 17:26 +0200, jmgonzalez@opensistemas.com wrote: > > > I''m trying to modify lustre cache parameters, but I don''t > find info. > > > The problem is that if I open a 1''5 Gb file to read in ext3, the > > > second > > time, > > > system can read it in 0.9 seconds, but if I open it in lustre, > > > second time > > > > > takes 11.2 seconds. It seems that lustre doesn''t use > right virtual > > > file > > system? > > > > > > Do you know which parameters should I must change?, I don''t find > > > under > > /proc > > > valid parameters to change it > > > > How large is RAM? There is /proc/fs/lustre/llite/fs0/max_cache_mb, > > which is default to 3/4 of RAM size. > > > > Cheers, Andreas > > -- > > Andreas Dilger > > Principal Software Engineer > > Cluster File Systems, Inc. > > > > > > > > > ---------------------------------------------------------------------- > Mensaje enviado por el servidor de openSistemas > (www.opensistemas.com) _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >