2010/6/23 Luc Chase <[email protected]>

> ez is based on an EAV data-model (allowing us to store any new
> content-class without changing the DB schema).  But it results in the the
> DBMS we currently use (e.g. MySQL, etc.) which is designed for more
> conventional relational database designs, having to work very hard to
> execute fetches.  KV stores are designed for systems such as eZ and are able
> to execute the same types of queries but more efficiently than a system
> which would have expected different content classes to be in different
> tables. The current situation is convenient and very good for a system that
> allows new classes at any time. But in terms of efficiency on the DBMS, it's
> like creating a dbms on a dbms.  So, using a KV store I would expect a
> performance improvement for fetches of maybe hundreds or thousands of
> percent. And scalability issues would be fully resolved.
>
To be fully honest, as a member of the eZ engineering team, i'm quite aware
of how eZ Publish is structured :-)

It is still a work-in-progress, but we are investing serious resources into
a more advanced content engine that would be able to use no-SQL databases
(sorry for the buzzword). We haven't investigated KV at all, but are
studying (and working on/with) CouchDB, MongoDB and more specifically Solr.

KV looks interesting though, but I confirm we haven't explicitely mentionned
it so far.

-- 
Bertrand
-- 
Sdk-public mailing list
[email protected]
http://lists.ez.no/mailman/listinfo/sdk-public

Reply via email to