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

Re: 2.1 & 2.2 statistics, and some odd behavior that needs to be examined.





--On Wednesday, February 11, 2004 3:07 PM -0500 Jong <jongchoi@OpenLDAP.org> wrote:

Any thoughts on why 'file-backed DB caching' is causing such a problem
under 2.2 as compared to 2.1?

Isn't it that 2.2 has higher write overhead than 2.1 and hence the 'shared memory DB caching' has higher impact than the 'file-backed DB caching' on their performances ? Assuming no swapping, continuous writes to a shared memory region can be performed without disk access, while writes to the named memory mapping can cause writeback of dirty pages if the number of dirty pages goes beyond a threshold. Since slapadding produces a lot of dirty pages, continuous disk write activities would be observed with the 'file-backed DB caching'.

Jong,

I'm doing a read based performance test -- No writing is happening.

Please see:

<http://www.stanford.edu/~quanah/directories/statistics/solaris/replica/per
formance.html>

read the description of the test, and look at the results.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/TSS/Computing Systems
ITSS/TSS/Infrastructure Operations
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html