Tuesday, October 09, 2012

Pluggable database and Peoplesoft architecture

After the announcement during the OOW2012, we know a very little about 12c, enough to make assumption though.
It is not coming this year, rather late next year, but one of the most important change will certainly be on architecture level, called "pluggable database".

What does it mean ?
Well, not very detailed yet, basically one database (the container, containing the core of Oracle data dictionary) and multiple databases on top of it (the pluggable databases containing application's code). It saves a lot of spaces, and over all a lot of resources.
That's a huge difference compared to the existing Oracle database architecture, one-to-one.

For the Peoplesoft people, it can remind similarity with existing Peoplesoft architecture (based on tables PSDBOWNER, PSACCESSPRFL and on users accessid and symbolicid).
For decade, we are able to install multiple application (HCM and FSCM for instance) in one single database. It shares the database binaries, engine, tablespaces... In the end, you have two separate applications with its own metadata, all in one database.
Good enough, but all is managed by the application code.
However, since each application has its own constraints and usage (one will be more batches oriented, the other more transactional), it was rather painful to tune the database properly. Hey, all is based on one and only one database engine, using same init parameters (it seems it will be different)... So, whether it is still possible within the latest Peoplesoft version, it is strongly a not recommended way of work.

Seeing that Oracle takes the way of consolidation to have multiple databases in one will be a real improvement for the Peoplesoft shop hosting several applications. Having more than one applications in a database won't be managed by the application as it was by the past (as said above), but by the database. Much stronger design, most likely more efficient, more secure and certainly more "tunable".
Especially nice when databases are already “connected” within messaging mechanism, probably more understandable.

The parallelism we can do between the existing People architecture and the coming database architecture, one database (the container), multiple applications (the pluggable databases), make me say that Peoplesoft was in advance... just not on the same level.

It's promising, really looking forward into it.



Noons said...

Indeed! Looking forward to it as well.
I run a lot of applications per database due to the nature of our servers and how they are setup.
For me this will be a major step forward - at last I see something really useful coming into a new Oracle release!
Let's hope it's done properly, with isolation at schema level and tunable per schema or per plugged db.
And no use of public synonyms - I want contexts used properly!

Nicolas Gasparotto said...

Hi Noons, there was that famous "segment deferred" feature also useful coming within 11gR2. Right, that was not a huge improvement as it could be with this one.
Cross finger and let's see exactly how it will be implemented.