Can you please clarify why you need the second set in the second data center.
I believe this is to mitigate the risk of loosing a data center. Kind of disaster recovery site. The requirement doesn't say that RAC instances must be in different data centers so they can run in one location. In fact, the requirement is either interpreted wrong or stated wrong (proposing EXACT solution instead of requesting SOME solution). Better solution using Oracle tools for remote datacenter replication would be physical standby. Alternatives, as Adar mentioned using third party replication.
1) You have one RAC node in each center, so each node will access the local copy.
This is not possible with ASM as there is no concept of "local" disks and remote disks. In ASM there is a concept of primary extent and secondary extent. ASM will try to distribute all primary extents equally across all disks in a diskgroup. Write is done to both places and read is always done from primary extent (unless it's unavailable). Assuming that your data is quite well distributed (which is the case with ASM) _AND_ all your disks deliver equal performance, this is a good approach. The latter "_and_" actially makes it quite a risk to put one of failure groups to remote datacenter as half of read operations will go to remote SAN box. Dynamic rebalancing for performance is NOT implemeted in ASM as some might be confused. It seems this confusion is quite common - automatic rebalancing in ASM is based purely on assumption that all data must be distributed as equal as possible and it doesn't take into consideration different performance characteristics of disks/SAN boxes.
> Hello All, > > We are planning on testing a two node 10g RAC using ASM for one of our > database on pSeries using AIX 5.3. We have a primary requirement of > mirroring data across TWO data centers (Hitachi SAN).
-- Best regards, Alex Gorbachev