On Tue, Jul 10, 2001 at 01:49:54PM -0400, Dan Melomedman wrote:
> I'd like to get an idea just how large some clusters people on this list
> running are. Anyone with 100,000 accounts and heavy loads? Just trying
> to get an idea how scalable qmail-ldap could be with it's forking model.
> Thanks.
Like Mike I don't have many accounts too, but aside from the great
scalability qmail-ldap offers it is still very performant. We run all mail
services mainly through one box (a second one is there for redundancy only,
an we use the clustering support for ease of routing to special machines),
an duron 650 with 384 MB RAM running OpenBSD. I didn't get the box to its
performance max whatever I tried, performance-wise the same rules as for
stock qmail should apply (the user lookups may be faster, but as with stock
qmail the queue is the limiting factor before). You wan't to make sure that
/var/qmail/queue is on its own 15k RPM SCSI disk or its own very fast disk
array, maybe even a solid state disk. Depending on you usuage type it may
make sense to put all Maildirs on its own disk(-array), too - if you do
mainly mailouts this gains nothing, of course.
--
* Henning Brauer, [EMAIL PROTECTED], http://www.bsws.de *
* Roedingsmarkt 14, 20459 Hamburg, Germany *
Unix is very simple, but it takes a genius to understand the simplicity.
(Dennis Ritchie)
