[Date Prev][Date Next]
RE: performance tuning issue
> -----Original Message-----
> From: owner-openldap-software@OpenLDAP.org
> [mailto:owner-openldap-software@OpenLDAP.org]On Behalf Of james liang
> I am doing some benchmark testing to evaluate openldap
> performance. The read access is very consistent and meets my
> need. However the insert performance is very disappointing. The
> insert new entry operation seems to depend on the size of the
> database linearly.
> I use the ldap server to store customer information. Each user
> gets it's own sub tree with about 100 entries. When I test the
> difference between a 500 users vs 5000 users, the insert operation
> (inserting a new user subtree) takes 8 times longer. I only have
> one index on objectclass. On my powerbook G4, the time went from
> 2 to 16 secs. I am using bdb backend. I understand openldap is
> for infrequent insert/updates. However, perf degradation of 8x
> when database size increases by 10x is still a very big problem
> for me. I've tried openldap 2.0.15, 2.1.3 and 2.1.4.
> I have the following question:
> What can I do to improve the insert performance?
back-bdb provides very few perf-tuning config options itself because the
Berkeley DB library provides quite a wide range of tuning features. You
should study the Berkeley DB documentation.
> Any compiler options to tweak?
Probably not, but every platform is different. Look in your platform's
> Any reading material I should look into?
> In general, any advise on tunning openldap?
It's generally accepted that oak barrels provide superior results for tunning
-- Howard Chu
Chief Architect, Symas Corp. Director, Highland Sun
Symas: Premier OpenSource Development and Support