We have been running our reporting tool (Cognos) against our Oracle
E-Business Suite schemas for some years now. This situation is about to end
and I am considering architectures that will lead me to a data warehouse
type environment. What I would like to consider is a logical standby
database running on a Linux based RAC. I'm reluctant to introduce RAC into
my environment without a fairly certain return on cost and complexity.
Could I use one node of the RAC to apply the logs to while leaving other
nodes free for reporting? Would this be a recipe for trouble given that one
node would be updating objects another node would be querying? Or could I
use one node to apply logs, another node to do ETL in the standby and yet
other nodes to report against the denormalized data?
Some of the criteria that are driving our discussions is the need to get
reporting off production fairly quickly and a readily scalable
architecture. I have heard that RAC is not all that scalable unless careful
consideration is given to the application so that use of the interconnect is
as small as possible. I know this question is very general but I would
appreciate any discussion or pointers to documentation or presentations.
We will be in a 10gR2 environment when this effort takes off.
Thanks
Allan
--
http://www.freelists.org/webpage/oracle-l
Received on Tue Mar 27 2007 - 09:19:10 CDT
