Skip to Content.
Sympa Menu

grouper-users - Re: [grouper-users] PSP performance

Subject: Grouper Users - Open Discussion List

List archive

Re: [grouper-users] PSP performance


Chronological Thread 
  • From: Tom Zeller <>
  • To: Rahul Doshi <>
  • Cc: "" <>
  • Subject: Re: [grouper-users] PSP performance
  • Date: Wed, 17 Oct 2012 10:43:02 -0500

> I provisioned roughly 2000 groups with 500000 membership using real-time
> changelog provisioning and it took about 6.5 hours. I then deleted all the
> groups from LDAP target and re-provisioned them using bulksync and it took
> just less then an hour. I am surprised by the difference in provisioning
> times using loader daemon versus bulksync. I am wondering if there is
> something I might have misconfigured in the real-time provisioning or it is
> expected behavior. I am running PSP 2.1.2.

Probably not a misconfiguration. Perhaps thinking of it this way will help :

Consider 500,000 ldap operations of the form

dn: cn=group,ou=groups,dc=example,dc=edu
changetype: modify
add: member
member: dn=cn=subject0,ou=people,dc=example,dc=edu
-

and less than 500,000 ldap operations of the form

dn: cn=group,ou=groups,dc=example,dc=edu
changetype: modify
add: member
member: dn=cn=subject0,ou=people,dc=example,dc=edu
member: dn=cn=subject1,ou=people,dc=example,dc=edu
member: dn=cn=subject2,ou=people,dc=example,dc=edu
...
-

Depending on the depth of your group nesting and average number of
memberships per group, the bulk operations should be faster when
provisioning everything. However, for a large group, bulk operations
are slower when provisioning a single membership (because the psp
reads everything when doing a diff). Directory performance may vary
based upon indexing and referential integrity too.



Archive powered by MHonArc 2.6.16.

Top of Page