[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: openldap bench / stress free tools

On Wednesday 08 March 2006 11:31, Jehan PROCACCIA wrote:

> About benchs, I've look rapidely through slamd, altough apparently a
> very powerfull utility it will take some time for me to make in run, I
> will probably when I'll get some time... ,

You can also use slapd-* from tests/progs to generate load. In my case, I 
wasn't that interested in the time taken to complete a search, but rather the 
number of searches, so I monitored performance via back-monitor.

> but for now I just timed 
> ldapsearch and I've seen lot's of performances disparities between my
> replicas , from 1m10s to less than 1s on a search on cn=*  -> 3715 entries
> .

Shouldn't your size limit for unauthenticated searches be lower?

> $ time ldapsearch -x cn=* -h openldap -b dc=int-evry,dc=fr > /dev/null
> real    1m10.303s
> user    0m0.424s
> sys     0m0.160s
> $ time ldapsearch -x cn=* -h ldap1 -b dc=int-evry,dc=fr > /dev/null
> real    0m0.709s
> user    0m0.500s
> sys     0m0.168s
> $ time ldapsearch -x cn=* -h ldap2 -b dc=int-evry,dc=fr > /dev/null
> real    0m25.819s
> user    0m0.452s
> sys     0m0.184s

Why are openldap and ldap2 so slow? Does the hardware differ so much? Or is it 
tuning (index, DB_CONFIG. limits)?


Buchan Milne
ISP Systems Specialist

Attachment: pgpQNMrQFKCzY.pgp
Description: PGP signature