Re: How to choose a DB

Miguel Cruz wrote:
>
> Anat Maoz <anatmaoz_at_hotmail.com> wrote:
> > I need to choose a database which serve my server.
>
> I have a similar need. I'm wondering if there are any good sites or other
> resources that do side-by-side comparisons of the major heavy-duty database
> platforms, so I can balance my recommendation using aspects I may not have
> considered in my own evaluations.

Unfortunately no such resource.  

> At the moment I'm looking most closely at Informix and Oracle because they
> are supported natively (sans-ODBC) by PHP. Key features for our application
> include transparent (or near-transparent) clustering and flexible fulltext
> indexing across selected fields.

The problem is that for both Oracle and Informix your two requirements, transparent cluster support and fulltext indexing, are both supported but by separate products. Oracle Cluster Server does not support DataBlades which would be needed for full-text indexing and searching and Oracle 8i does not support clusters. Similarly Informix Extended Parallel Server, a more advanced product that OCS incidentally, does not support DataBlades either and Informix IDS.2000 does not support spreading a database/table across a cluster the way that XPS can. Informix is reportedly working to merge the XPS and IDS code bases in the same way that the recent release of IDS.2000 merged the code base from the basic IDS engine with that of the Informix Universal Data Option product to create a Universal Server that has the features and performance of Informix's full blown OLTP engine (actually it is reported to be faster). So you will have to decide what features you want most.

Actually, it is my understanding that OCS only adds the ability to share disks between cluster members and actually implements spreading a table over cluster members using SYNONYMS and VIEWS UNIONing the various pieces of the table back together. This is in contrast with Informix XPS which is shared nothing (except when one of the cluster members crashes) and actually treats the database as residing on ALL members of the cluster invisibly querying all co-server table fragments in parallel. There is no reason one could not manually implement the Oracle style pseudo-coserver scheme using IDS.2000 or Oracle 8i and take advantage of those servers more advanced data support.

Art S. Kagel Received on Wed Dec 29 1999 - 00:00:00 CET