SUN DSEE 6.2 vs Fedora DS 1.1 performance comparison

Question

Hi all,
I've just discovered a nice tool from SUN about performance analysis for ldap servers named SLAMD (http://www.slamd.com)
So I configured it and tried to analyze my servers. I've setup one SUN DSEE 6.2 and one Fedora DS 1.1
in my workstation. Both of them being populated with the same data (160 sample entries from sun) and using the same file descriptors.
My workstation is running fedora 8, Core(TM)2 Duo CPU E6550 @ 2.33GHz / 2 GB ram.
I did a couple of tests but all of them had the same search filters
Entry DN           ou=people,dc=example,dc=com
Search Filter           objectClass=*
Attribute(s) to Compare/Modify      Add Operation Frequency           3
Compare Operation Frequency           7
Delete Operation Frequency           4
Modify Operation Frequency           4
Modify RDN Operation Frequency           1
Search Operation Frequency           10      description
I will give the results of my final test which lasted 240 seconds / 200 threads from one client
DS Overall Operations (Average/sec)
SUN *35,858*
Fedora *304,867*
It seems to me there is a huge difference! I didn't expect to get such numbers. To tell you the truth
I expected SUN DS to be much faster that Fedora DS instead of being *10 times slower*.
Furthermore while running the test on the Fedora DS the system got a max load of around 7-8 which implied that the system
worked hard to perform the test (CPU always at 100%).
On the other hand while running the SUN DS test, the system never got load more that 1 (cpu not more that 22%).
It was like the SUN DS was capable to do better but it was never bothered. I played with indexes, file descriptors, number of threads without
any significant change of performance.
I'm sure SUN DS can do better. So I'm looking for thoughts on the subject as well as performance tunning/optimization documentation.
Is the resource kit also available for 6.2 or is it just for SUN ONE server?
regards
Giannis

Answer 1

Giannis,
Giving raw performance numbers doesn't mean anything unless you also provide the details of the data in your directory server, the settings and the exact tests performed (if it's a slamd standard job, give its name).
Slamd contains many jobs that are doing many different things leading to completely different numbers in term of operations per second.
This said, the numbers you show are puzzling me : SUN 35,858 vs Fedora 304,867 (Operations / Second) ?
I assume the , is the unit separator (and not like in the US the separator between thousands and hundreds).
If so, there is definitely something badly configured on Sun DS and/or Slamd.
Regards,
Ludovic.