On Sun, 17 Jun 2001 09:53:28 -0700 Chuq Von Rospach <chuqui@plaidworks.com> wrote:

On Sunday, June 17, 2001, at 09:39 AM, J C Lawrence wrote:

...

Exim for instance will defer deliveries if more than N messages are received in a single connection. As a result, you typically get no outbound deliveries going on during a qrunner broadcast.

IMHO, I consider that seriously broken.

Its configurable.

...

I've done similar experiments with news spools. The values are incredibly subjective to RAID stripe size. Bob Mende IIRC did some interesting work here which I think he presented at LISA.

FWIW We ended up with a 1Meg strip size as the sweet spot, much bigger than we'd expected.

I know someone who did the same, and found the best perforamnce iwth (I kid you not) 32 meg strip sizes.

Not surprised. Among other things its going to be sensitive to number of spindles, RAID type, and feed characteristics.

...

Silicon disks are another one. At Critical Path we used solid state disks for /var/spool and Qmail fairly, umm, whirred.

But what I've found, for really large e-mail installations, there's always another bottleneck. The bigger/faster machine paradigm just doesn't scale after a while, so what I'm working on now is a new setup that I'm calling the "army of smurfs" design. I'm going to be buying lots of small/fast/cheap boxes, and not going to try to try to keep making that single monolithic machine do incrementally more.

I've noticed that E-Machines are getting popular for precisely that use/need.

-- J C Lawrence claw@kanga.nu ---------(*) http://www.kanga.nu/~claw/ The pressure to survive and rhetoric may make strange bedfellows