Is the cacheSize value defined for the partition has any influence on the
cacheSize values defined for the individual attributes?
Those cache are different (how smart is it to have defined different
caches for attributes and entries is questionned atm ...).
- Attribute caches are used to cache the attribute's index
- Partition cache is used to cache the entries themselves.
I would say that dependening on your filter, both caches will have
different impact on performances. If you consider a standard LDAP server
(whatever standard means ...), you will have more disk space dedicated
to index than to data, so a good balance should be done between the
cache size for each kind of elements (Attributes of Entries).
Again, if you want optimal performances, cache all the entries and
attributes (set 40 000 for every cache).
NOTE : If you store HUGE data into entries - like JpegPhoto -, then this
strategy can be totally wrong...
Another point : when searching for an entry, using such a filter :
(uid=acme), the server will access to the uid index, and then t the
master table, containing the entries. Caching will help in two ways :
less index will be loaded from disk, and when the entry 'pointer' will
be found, as it may be cached, less disk access will be done when
reading the entry.
Last, not least : cache is loaded on the fly, so if you want to observe
increased performance, you will have to wait for the cache to be loaded,
ie unless you already have read data from disk at least once, it's still
on disk and then, slow to get :)
--
--
cordialement, regards,
Emmanuel Lécharny
www.iktek.com
directory.apache.org
