RE: ber_memalloc size distribution of back-ldbm

["Jonghyuk Choi" <jongchoi@us.ibm.com>]

The directory data set was synthesized using DirectoryMark toolset.
I'll mail it to you if somewhat large size is ok.
I also assume that there may be a commonly used data sets in the community
for this purpose... then I'll use them.

What slab allocator does is to reduce the number of calls to buddy system.
These two methods can be jointly used and  be additive with each other
in improving performance.

- Jong

jongchoi@us.ibm.com
IBM Thomas J. Watson Research
Enterprise Linux Group


"Howard Chu" <hyc@highlandsun.com> on 2001-10-25 10:24:07 PM

To:   Jonghyuk Choi/Watson/IBM@IBMUS, <openldap-devel@OpenLDAP.org>
cc:
Subject:  RE: ber_memalloc size distribution of back-ldbm



Can you make your 100K entry data set public? It would be nice if we all
had a consistent data set available to measure performance against.

Certainly a fast malloc library would be a good thing, but a better thing
would be to eliminate unnecessary use of malloc in the first place. As I've
already demonstrated, this is not too difficult to accomplish on a case by
case basis.

  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc