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

Re: OpenLDAP 2.1 (?) on RedHat Enterprise summary

> On Fri, 7 May 2004, Pierangelo Masarati wrote:
>> Any details on the configuration that might be significant?
> No, almost everything is identical between both test "systems"
> (single machine two seperate identical SCSI disks used for the
> "systems").
>> e.g. I assume the same underlying database was used, and the
>> same indexing was present, and there was no extra load on the
>> machine, and the cache size was the same in both cases, and
>> the same cache initial status was obtained.
> Yep, and the same database files themselves were used too.
>> I'm testing with latest 2.1, 2.2 and HEAD code, for obvious
>> reasons; stock rpms are not a choice.  I'm going to run similar
>> tests on both distributions ASAP.
> Well, the results I gave you were the ones RH wanted, I did tests on
> 2.1.25/db- as well which exhibited the same behaviour.

I had a chance to run one of our internal tests, but unfortunately I don't
have exactly the same hardware for both AS2.1 and AS3.0; slightly
different disks, same processor (Xeon 2.6Ghz HT) but one on the 2.1 and
two on the 3.0.  I generally had better peroformances on 3.0 than on 2.1
on tests that should not be affected by the number of CPUs.  Typical
performances were 8 min vs. 11 min. to load 110000 entries (inetorgperson
with cn,sn,givenName,uid,mail,preferredLanguage,userPassword, most of them
with substrings indices)  but in one case, on 3.0, I had to use a larger
cache in DB_CONFIG (1Gb vs. 512Mb); 4 s vs. 7 s to dump the entire
database (in cache!) with ldapsearch as rootdn.  I need to turn these
tests into something more scientific, but the increase in performance is


Pierangelo Masarati

    SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497