At 02:43 PM 4/3/2002 -0800, Joel Laing wrote:
>I think the biggest improvement you can make is to put the temp-drop-dir 
>on a separate drive. This will help lessen the I/O bottleneck. Also, use 
>fast drives. If possible, stripe multiple drives for performance on both 
>the spool  file system, and the tmp-drop-dir file system. Mind you, I've 
>not really played with the 4.x possibilities, but the above greatly 
>improved performance for my 3.x setup.

No necessarily.  If the temp file and maibox are on different
partitions, qpopper had to copy the temp data back to the
mail box.  If they are on the same partition qpopper can (and
does) do a move (unlinks the old file, relinks the new).

For a small spool, I found a separate disk faster.  For our
current spool having the temp files on the same partition is
much faster.  (Lookup fast-update option in qpopper manual).

You should also consider using a hashed spool, which decreases
lookup times in directories, and avoids lock contention.  Server
mode also uses the .cache files to avoid scanning large mailboxes
when they do not have new mail---this is a big win.

Now, the $64,000 question. The .cache code is incomplete.  The
cache file stores the header information qpopper returns to the
client.  If the .cache file is older than the mailbox, qpopper re-scans
the mailbox.   This is not necessary, the information in the .cache
file is still good, it is just incomplete.  It would be much (much (much))
faster to use the .cache information to scan only the new mail.

I could see two options, a --fast-cache and a --safe-cache.  Fast cache
would assume that nothing else affects the mailbox file, and all new
mail is appended.  It would just read the .cache and seek to the next
new message.  --safe-cache would assume that something else 
might remove messages (such as an old message scanner), and so
would seek-and-confirm each .cache entry.  This would still be faster
than a re-scan.   The source says that .cache is incomplete.  Is anybody
else completing it?

Regarding scanning the spool and removing old mail--this is a very good
idea.  We do this once a week, it typically removes 600-800meg of old
mail.  Without it, we would soon be buried in old messages.  It would
be nice if qpopper could do this. Say an option to remove messages
that are more than n days old.

Mike


Another 
--
Michael Sofka                          [EMAIL PROTECTED]
CCT Sr. Systems Programmer  email, webmail, listproc, TeX, epistemology.
Rensselaer Polytechnic Institute, Troy, NY.    http://www.rpi.edu/~sofkam/

Reply via email to