[Tfug] a comparison of file systems

[t takahashi]

true.

what i had in mind, however, and thought some obsessive geek or
file-voracious company might have already done for its own purposes,
is to catalog fs reliability differences in a table, and unless it's a
huge competitive secret, publish the results.  even if it's
library-dependent.

but i really meant reliability at the fs level, not performance or
limitations of apps.  just > >> < lseek, whatever.  do stuff like try
lots of filename lengths, and all inodes, and 10 million files in a
directory, and create races, and simulate errors, and vary levels of
fragmentation, repeatedly and with varying load conditions for a
month.

i'm not upset that nobody has done it, just bemused.  features are
less important to me than reliability.

(pardon the top-post; i cannot edit headers to provide context.)

On 11/8/05, Robert Hunter <bobjim.hunter at gmail.com> wrote:
> Yes some performance stats would be nice.  But such tests would be
> influenced by more factors than just the raw capabilities of the file
> system ( i.e. architecture, OS, libraries )    For instance, just
> because NTFS theoretically allows path names to be several thousands
> of characters in length doesn't mean you can create or use such a file
> with common applications in Windows.  This is a limitation of win32,
> and not the underlying file system.
>
> On 11/8/05, t takahashi <gambarimasu at gmail.com> wrote:
> > yes, it is nice.
> >
> > what i found harder to figure out when i looked into it a few months
> > ago is which fs is most reliable.  it seems to be a closely-held
> > secret, which means that most people have an opinion that is not worth
> > listening to.  the general opinion is in favor of ext3 fwiw.
> >
> > surprising that intense stress testing hasn't been done with the
> > degree of thoroughness of that table.  i doubt that
> > xfs/jfs/rieser/ext* would all pass such testing, if it were done
> > severely and even moderately cleverly for a month.
> >
> > i use ext3, but only reluctantly, and might try xfs at some point for
> > large files.
> >
> >
> > On 11/8/05, Robert Hunter <bobjim.hunter at gmail.com> wrote:
> > > This is such a nice reference that I felt the urge the spam the list
> > > with it. :-)
> > >
> > > http://en.wikipedia.org/wiki/Comparison_of_file_systems
> > >
> > > --
> > > Rob
> > > _______________________________________________
> > > tfug mailing list
> > > tfug at tfug.org
> > > http://www.tfug.org/mailman/listinfo/tfug
> > >
> >
> >
> > --
> > Webmaster: do you believe that people will switch browsers to view
> > your page instead of going to your competitor?
> >
>
>
> --
> Rob
>

--
Webmaster: do you believe that people will switch browsers to view
your page instead of going to your competitor?