> I'll second that ... I've never gotten a Unix box of any flavor above 7%
> fragmentation and most of that was log files. But as soon as logrotate

Well, given that log files can be fairly large, that more or less
makes sense, that there might be some fragmentation. But how would you
measure it on modern systems? For instance I seem to recall a util for
ext2fs that would tell you how fragmented a file or filesystem was and
even some versions of 'mount' would tell you the fragmentation 
percentage of a given partition).

Also because log files are loarge, isn't some of that fragmentation 
already "engrained" into the file because of indirect blocks and so
forth? (Maybe true for ext2/ext3 - I don't know if even reiserfs or
other newer filesystems use those) but for what it's worth, any file
of large enough size is going to have extra blocks for pointers and
such.

> Civileme talking about some guy in Alaska who managed to hit some unreal
> number like 90% fragmentation... but I don't know much more than that.. 

That's pretty awesome :). But I agree that adding more RAM or having
a faster disk is a good thing to do. I would think that adding RAM is 
going to make more of an impact over other upgrade paths. 

> There might be a way to further optimize a file system, but this is
> pretty much beyond the ken of what I know how to do.  For me.... Get a

Well, there might be ways to do background reshufflings of files in
the filesystem dynamically -- if based on usage (i.e., files accessed
more recently or more often are the ones that the OS or filesystem
pays more attention to). After all, if files aren't accessed all that
often, why bother optimizing them at all?



Want to buy your Pack or Services from MandrakeSoft? 
Go to http://www.mandrakestore.com

Reply via email to