RE: Questions...

["Howard Chu" <hyc@highlandsun.com>]

> -----Original Message-----
> From: Jong [mailto:jongchoi@OpenLDAP.org]

> > Argh, spoke too soon. HEAD is several times slower than
> RE22 now. I'll
> > probably revert the idl.c changes until I figure out what to fix...
>
> Is that the indexing is not fast than observed previously or
> other operation's performance
> such as searches is affected by the idl caching change ?

Indexing is not as fast as I previously observed; the previous results were
not consistently reproducable due to other bugs. When an IDL is not present
in the cache, the new code is significantly slower than the old because it
reads the entire IDL (which may be large) before making any change.

The old code changed the on-disk IDL after reading only a few slots. It
appears that this approach is still the best. Given the large number of keys
in just the dn2id index, it seems impractical to set the IDL cache size large
enough to be of use for a large number of adds.

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