ext3 fsck vs other filesystems

Broekman, Maarten Maarten.Broekman at FMR.COM
Wed Oct 28 14:21:51 UTC 2009


I've been asked by my manager to perform some performance testing on the
fsck performance on ext3 vs other filesystems (such as vxfs).  In
particular, we're trying to look for the point at which we can
definitively say ext3 doesn't cut it.  This has come about because the
performance of fsck on some of our larger ext3 filesystems takes upwards
of 8 hours, which isn't acceptable in our production environment.
However, because of the licensing costs, we don't really want to say
that all filesystems on SAN storage need to use a filesystem other than
ext3 since that would mean even the small filesystems have to incur the
licensing costs.

Can anyone either point me to some whitepapers that talk about this?
Alternately, can anyone make some recommendations on the best way to
test this?  I've been doing some rudimentary tests, but the fsck times
look ridiculously short (15 minutes for a 600GB ext3 filesystem)
compared to what we've seen in production (12 hours for a 600GB ext3
filesystem).  What influences the time an fsck takes?  That may give us
some ideas for restructuring the data in the filesystem as well.

Thanks
Maarten Broekman 
Email: maarten.broekman at fmr.com 






More information about the redhat-list mailing list