Re: 9i streams vs triggers

Steve Jorgensen wrote:

> On Thu, 09 Jan 2003 22:16:44 +0100, Sybrand Bakker
> <gooiditweg_at_nospam.demon.nl> wrote:
>
> >On Thu, 9 Jan 2003 13:06:43 -0600, "servant" <mjohns1_at_uhc.com> wrote:
> >
> >>I have been tasked with developing a reporting architecture for an
> >>application that has been developed over the last few months. I have been
> >>told there are 2 main concerns: performance (contention) and complexity of
> >>the data model. I am thinking of using Oracle's replication to maintain a
> >>(near) real-time replica. That (basically) resolves the contention issue.
> >>The complexity issue is more difficult.
> >>
> >>I am thinking about using 1) streams against the redo logs or 2) triggers to
> >>pull the data from the replica, apply business logic to it, and populate a
> >>"logical" data model on which reports can be generated. I would like to use
> >>the same abstraction layer to go from the physical to the logical data model
> >>as the application interface uses (Java objects with EJBs). In other words,
> >>I need the code inside the streams or triggers to call Java methods (ie:
> >>loadCustomerData(cust_id)) which would get the data and apply any business
> >>logic and then populate the "logical" database.
> >>
> >>In essence, I think I am being asked to create a data warehouse, but I
> >>haven't convinced myself of that yet.
> >>
> >>Anyway, here is my question: Does anyone have an opinion on which method
> >>(streams or triggers) would tax my replica database more--from a CPU,
> >>network, and/or contention perspective? Any opinions from other
> >>perspectives which option is better (maintenance, etc).
> >>
> >>
> >
> >First of all I would like to question the need of a replica database
> >for reporting purposes. As for contention, this is simply not an
> >issue, if the application has been designed properly. Your post has a
> >distinct smell of finding a 'hardware solution' read 'workaround' for
> >what is in reality a software design problem.
>
> That's odd because everything I've read, including a book specifically
> on Oracle DBA suggests the schema designs and optimal database
> configurations for real-time and analysis are diametrically opposed,
> so you always have to choose either a compromise between the 2 or use
> separate databases and servers.

Just a note for the odd person that gets a stern response from us such as 'RTFM' and reacts negatively.

Take note of how we have reacted to servant's request for help.

Try to figure out why we are being so helpful here and not in other cases.

We have repeated this many times but it bears repeating again. We gladly jump in and help those who don't come to us with an attitude that they have 10+ years of experience blah blah blah but rather help those who are willing to dig in and read, learn, and help themselves.

Daniel Morgan Received on Fri Jan 10 2003 - 16:07:43 CST