Jens-S. Voeckler wrote:
> Therefore, I would be utterly interested how your beast behaves after your
> cache has been full for a few days. It might not be as beneficial as it
> looks now, but I may be off the track.
There is much more to it in reiserfs than only journaling. The
filesystem is very different from your average UFS derivate (including
Linux ext2). The speed benefits in reiserfs comes mostly from the
different filesystem design, and not from the journal. The filesystem
without a journal runs quite a bit faster, but they have selected to
only provide the filesystem with journaling at this time.
You are correct in that a filesystem journal only gives the disk more to
do, and for a Squid with async-io the journal by itself is likely to
only slow things down, except for large metadata changes. However, for a
Squid without async-io a journal can speed things up by allowing Squid
to perform certain operations without blocking since the journal allows
the filesystem to run with other constraints on metadata updates. This
performance effect can be acheived by other means such as linux async
filesystem mode (the default), or fastfs on Solaris.
Received on Wed Dec 22 1999 - 06:25:57 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:50:05 MST