directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Johan Asterholm" <asterh...@gmail.com>
Subject Re: Partitions
Date Mon, 22 Jan 2007 14:38:06 GMT
Hi,

By backend can store up to several milion records but its quite fast in
answering (about ~20 MS) when having 2 milion records. The application will
handle caching itself so no caching wold be necessary. Basically what I'm
looking for is just an example for writing my backend.

BR / Johan Asterholm


2007/1/22, Emmanuel Lecharny <elecharny@gmail.com>:
>
> Hi,
>
> the backend is using its own structure atm. It's based on JDBM, which is a
> B-tree storage.
>
> The DN are also stored in a B-Tree, and this is not necessarily optimal.
> We may want to go for a H-tree instead of a B-tree for these DN (B-tree and
> H-tree storages are available in JDBM)
>
> So basically, it's just a question to associate a H-tree to the DN
> storage.
>
> Ok, this is theory. In the 'real world', this would need more analysis,
> and we must evaluate the impact of such a modification on the code base.
>
> Now, I have a question : how big is your data set ? If it's a few
> thousnads, there would be a very little increase in performance doing that :
> you have to consider that the cache system will very quickly keep in memory
> all the DN of your database. (you might want to increase the cache size for
> this kind of objects in your partition :
> ...
>    <property name="indexedAttributes">
>       <set>
>         <bean class="
> org.apache.directory.server.core.partition.impl.btree.MutableIndexConfiguration
> ">
>           <property name="attributeId" value=" 1.3.6.1.4.1.18060.0.4.1.2.1"
> />
>           <property name="cacheSize" value="100" />
>         </bean>
> ...
>
> In this default configuration, 100 DNs are kept into the cache. Just
> increase it to fit your needs.
>
> If in cache, grabbing a DN is really fast, wether you use H-tree or not.
>
> You will just have to consider if this would be of interest to modify the
> backend (with all the burden it suppose) instead of increasing the cache.
> And this is again a question of how many DN will you have.
>
> I hope this help.
>
> Feel free to go further, we really have this Hashing in our radar.
>
> Emmanuel
>
> On 1/22/07, Johan Asterholm < asterholm@gmail.com> wrote:
> >
> > Hi,
> >
> > I'm trying to develop LDAP integration towards an application, but I
> > have a hard time to understand how it all should work.
> >
> > To learn how it should be done I'm trying to develop a backend which
> > works towards a HashMap where the hash Map includes a key (Identifier) And
> > then another hashmap with attributes (They are not "Attributes" but simply
> > String names of the attribute). I prefer this way since this is quite
> > similar to what i need to do when integrating towards or application.
> >
> > so searching for  uid=1,ou=user,cn=example,cn=com would point to a row
> > in the hashmap, but ou=user,cn=example,cn=com is static.
> >
> > Is there any example where a simple backend like this is done? Maybe not
> > with a HashMap but something similar?
> >
> > This would help me tremendously.
> >
> > BR / Johan Asterholm
> >
>
>
>
> --
> Cordialement,
> Emmanuel L├ęcharny
> www.iktek.com

Mime
View raw message