On Fri, 19 Nov 1999 18:53:00 -0500 (EST) , [EMAIL PROTECTED] writes:
> Doesn't a journeling FS incure speed penalities due to the fact that all
> disk activity is logged?
Theoretically, a journalling FS should be able to get a better
response time on an fsync() call than an FS with soft updates, or
sync metadata. This will increase qmail's performance, since each
message incurs 4 or 5 fsyncs (two from qmail-queue, and two or
three from qmail-send during preprocessing).
Which reminds me, in the qmail FAQ, Dan says not to use qmail on
a filesystem with soft updates. I don't see why not, as long as
the soft updates implementation honors fsync() (which McKusick's
does). Am I missing something, or was the statement directed toward
users of more experimental softupdates implementations (e.g. Ganger
and Patt's)?
--
Chris Mikkelson | "Unfortunately, simplicity is a complicated mess
[EMAIL PROTECTED] | of a concept." --Taner Edis
