[PATCH] Cache messaging dgm connections

[Volker Lendecke]

On Sun, Sep 11, 2016 at 10:32:22PM +0200, Ralph Böhme wrote:
> hopefully better, I don't have any numbers. :)

Before and after the patch. 333823 vs 703036 messages per second.

Volker

vlendec at vl-vm1:~/git/scratch$ bin/smbtorture //foo/bar -U% local.messaging --opt
ion="torture:timelimit=10"
smbtorture 4.6.0pre1-DEVELOPERBUILD
Using seed 1473657374
time: 2016-09-12 07:16:14.804927
progress: 1
test: ping_speed
time: 2016-09-12 07:16:14.806171
Sending pings for 10 seconds
waiting for 20 remaining replies (done 1669114)
sending exit
ping rate of 333823 messages/sec
time: 2016-09-12 07:16:24.816016
success: ping_speed
vlendec at vl-vm1:~/git/scratch$ bin/smbtorture //foo/bar -U% local.messaging --option="torture:timelimit=10"
smbtorture 4.6.0pre1-DEVELOPERBUILD
Using seed 1473657439
time: 2016-09-12 07:17:19.456846
progress: 1
test: ping_speed
time: 2016-09-12 07:17:19.457756
Sending pings for 10 seconds
waiting for 20 remaining replies (done 3515250)
sending exit
ping rate of 703036 messages/sec
time: 2016-09-12 07:17:29.458367
success: ping_speed
vlendec at vl-vm1:~/git/scratch$

> 
> > I ask because Douglas did a very similar analysis of some 'obvious'
> > performance patches I did:
> > 
> > https://www.samba.org/~dbagnall/perf-tests/ldb.html
> > 
> > Likewise, when I added a cache of LDB index values between
> > transactions, it was obvious that it would help, because an index read
> > would be avoided, except when Douglas and I checked the performance, we
> > found the cache management costs took up the same space.
> 
> The cache is merely a list of active connections to messaging peers,
> so searching a cached connection means traversing a list which will
> typically contain just a few elements.
> 
> Cheerio!
> -slow
>