Re: Er.... for Sendmail users......enough

From: Christopher E. Brown <cbrown@woods.net>
Date: Tue Sep 12 2006 - 00:16:33 AKDT

On Fri, 8 Sep 2006, Matthew Schumacher wrote:

> Christopher E. Brown wrote:
> > On Thu, 31 Aug 2006, Matthew Schumacher wrote:
> > As to performance, last mockup I did was a few years back, with systems
> > running the same message load on the same hardware/OS (Quad Xeons, 2G
> > RAM, hardware RAID5 USCSI, Slackware, 2.4.x kernels).
> >
> >
> > Overall was about the same, with message lookup being very slightly faster
> > on Cyrus for accounts with < 1000 messages, and moderatly faster with
> > 10,000 messages. Message deliver was the reverse, with Courier (maildrop
> > LDA) running 15 messages/sec faster (35 vs 20) system wide. Both system
> > were doing Quota enforcement. And while Courier deliver performance
> > stayed nearly the same as the message count in each account grew (forced
> > quota recalc every 15 min slowed a bit), Cyrus dropped to 12 messages/sec
> > once the average message count per account topped 5,000.
>
> Hmmm, perhaps it's time to do another mockup. I'm wondering if this is
> still the case given that benchmark I posted. Cyrus is very fast, and
> if it does double couriers performance such as the benchmark suggests
> then hot standby cyrus servers are a wash performance wise, although I
> do agree that clustering is a bit simpler.
>
> I really need to do some benchmarking soon.....
>
> schu
> ---------

I have only seen the "twice native maildir" performance on an otherwise
unloaded system. Once the systems were going IO or CPU bound (my
benchmarks were run with 25, 50 and 100 IMAP clients hitting the systems,
listing messages, moving blocks of messages from folder to folder, etc)
the faster lookups remained, but message access itself was the delay. On
deliver/move the indexing was just more file ops to be done, and more CPU
cycles.

I do want to note that both system had enough memory allocated to fs
data/metadata buffer cache that the entire active directory structure was
more or less in memory. This would tend to make the directory walks in
straight maildir less expensive than an untuned system since 95% or the
required info would normally be in cache.

---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.
Received on Tue Sep 12 00:17:05 2006

This archive was generated by hypermail 2.1.8 : Tue Sep 12 2006 - 00:17:05 AKDT