jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andrei Latyshau" <andrei.latys...@gmail.com>
Subject Re: Dynamic Nodes
Date Fri, 08 Aug 2008 13:22:17 GMT
Thank you very much for information! I prepared XML example (at the end)
illustrating the hierarchy of objects and added comments with orders of
counts of objects.


MyObjects and Sub-Objects within MyObject can have different types -
different set of properties. Sub-sub-objects have the same type within
sub-objects.


1) "MyObject"s will be searched by simple queries with "header" but also may
be by fields from nested sub-objects.

2) Sub-objects and Sub-sub-objects can have attachments.

3) Sub-objects will be retrieved with filtered collection of
sub-sub-objects.


Do you find such hierarchy productive ? Or better to use usual relational
database?


<RootNode>

  <OrgPath...Path>

    <MyObjects>

      <!-- MyObject x 100 000 - 1 000 000 - can be archieved -->

      <MyObject>

        <Header Attr1="" Attr2="" Attr3="" />

        <Collection>

          <!-- SubObject x 10 -->

          <SubObject AttachmentsCanBeHere="">

                   <Header Attr1="" Attr2="" Attr3="" />

            <Collection>

              <!-- SubSubObject x 20 -->

              <SubSubObject AttachmentsCanBeHere=""></SubSubObject>

            </Collection>

          </SubObject>

        </Collection>

      </MyObject>

    </MyObjects>

  </OrgPath...Path>

</RootNode>





2008/8/8 Ard Schrijvers <a.schrijvers@onehippo.com>

>
> > > Is it possible
> > > to run JackRabbit on Oracle to guarantee good perfomance and
> > > reliability on 10-100 mio records?
>
> > Marcel Reutegger wrote:
> > I've seen jackrabbit work well with about 10 million nodes.
> > does anyone else have experience with larger repositories?
>
> IMO the relation between number of nodes and good performance is highly
> correlated to *how* you want to access these nodes and *how* the
> structure of these X mio nodes are (ie, not > 1000 direct childs per
> node). If you want to access nodes by simple getNodes() and 'easy'
> queries, I saw numbers of > 100 mio nodes on this list. If, though, you
> also want very complex CPU intensive searches (like searches with
> initial path constraints or many wildcard queries) you might hit a limit
> much earlier.
>
> -Ard
>
> >
> > regards
> >   marcel
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message