[syslog-ng]performance test questions

Roberto Nibali syslog-ng@lists.balabit.hu
Mon, 24 Feb 2003 13:03:05 +0100


Hello,

> I'm not really surprised the macro expansion has performance problems. When
> I wrote that code only 6-8 macros existed, I've checked my table and it
> contains 51 entries. All looked up sequentially. Very bad performance wise.

Indeed.

> Attached you'll find a patch which changes this to a faster algorithm. I
> have also tested it.

I'll give it a spin.

> Our new webmaster wanted to remove the old 'apache directory index' pages
> specifically. You can grab the tree with wget however.

I'll add it to our mirror script.

>>>Can you check if it works for you?
>>
>>I'll compile and beat it tonight.

I didn't get to test it out yet but I just started now. I hope you haven't 
released 1.6 yet :)

> The new snapshot to be generated tomorrow at 6AM CET will contain the patch
> I've attached now.

I'm working on with syslog-ng-1.6.0rc1+20030224 and libol-0.3.9 (nothing 
important seemed to have changed there). I will report back on Wednesday. So far 
it looks pretty good:

syslog-test# date
Mon Feb 24 12:59:47 CET 2003
syslog-test# cat /proc/$(pgrep syslog)/status
Name:	syslogp
State:	S (sleeping)
Pid:	30398
PPid:	1
Uid:	0	0	0	0
Gid:	100	100	100	100
Groups:	
VmSize:	     676 kB
VmLck:	       0 kB
VmRSS:	     524 kB
VmData:	     216 kB
VmStk:	       8 kB
VmExe:	     436 kB
VmLib:	       0 kB
SigPnd:	0000000000000000
SigBlk:	0000000000000000
SigIgn:	8000000000001000
SigCgt:	0000000000014001
CapInh:	0000000000000000
CapPrm:	00000000fffffeff
CapEff:	00000000fffffeff
syslog-test#

My first tests indicate a template speedup of 13% for 3 clients using UDP. I 
will do more indepth tests on Wednesday morning.

Thanks and best regards,
Roberto Nibali, ratz
-- 
echo '[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc