read the message again, specifically the part about "non-volatile."
this is starting to get pretty far off topic for this list, but i'll offer a
couple of observations:
software raid is pretty much by definition slower than writing straight to
disk for a single block at a time, because you still have to write the same
data out but you also have to do the bookkeeping involved. as you start to
write multiple blocks, you can start to see improved performance as you get
rid of a single bottleneck, but at the low end it's a waste.
since we're talking about the qmail queue dir here, software raid is pretty
silly for the queue dir. messages tend to be small - on our isp mail
servers, something like 70-80% of all messages are less than 10k. software
raid is useful to lump multiple identical disks into a single large store
that spreads the load evenly across spindles, but it's really more of an
administrative enhancement than a performance enhancement. we use sw raid
for the actual mail spool storage, and local uw scsi disk for the queues.
hardware raid is a different beast entirely, particularly when you're
talking about stuff that has battery backed caches. i still don't know that
it would be a win for qmail unless you had a LARGE queue with lots of large
messages.
shag
=====
Judd Bourgeois | CNM Network +1 (805) 520-7170
Software Architect | 1900 Los Angeles Avenue, 2nd Floor
[EMAIL PROTECTED] | Simi Valley, CA 93065
Quidquid latine dictum sit, altum viditur.
----- Original Message -----
From: Florian G. Pflug <[EMAIL PROTECTED]>
To: John White <[EMAIL PROTECTED]>; qmail mailing list
<[EMAIL PROTECTED]>
Sent: Mon 22 Nov 1999 18.04
Subject: Re: disk mirroring
> > I don't really mean to be a hardass here, but you need to know about
> > how the qmail queue works. You have the qmail source right? Included
> > with that source is an "INTERNALS" document which describes how the
> > queue works. With qmail's insistance on fsync'ing, you can see how
> > a writeback cache on the HW RAID controller can help.
> >
> > Or perhaps you don't know? HW RAID controllers can come with
non-volitile
> > RAM caches. When part of this cache is in "writeback" mode, scsi write
> > commands are put in the cache, and the controller tells the OS that the
> > command has been completed. Then the writes are committed to hard drive
> > (which have their own caches). Thus, multiple small-block writes
followed
> > by fsync's should finish much quicker on a HW RAID with writeback cache.
> >
> > If you're relying on OS RAM to do the same thing for a filesystem, then
> > the fsync will put an end to that.
>
> All this would make hw-cach *forbidden* for qmail queue dir, since then it
> is *not* guaranteed, that what is synced is writted on disk and will
> survive a power loss....
>
> Anyway, what is noone mentioning raid 5? I just played with it under linux
> (software raid) until now - but it seems quite fast.
>
> Greetings, Florian Pflug
>
