We have Oracle Dataguard replicating the master database across two data centers, a master in each DC, one primary, the other backup with the fault monitor taking care of failover.
Its not a seamless failover. Some DBA intervention is required to fail the database over but its relatively quick. Our data centers are too geographically dispersed to run a full RAC cluster which would be the preferred option.
There may be alternative methods such as SAN-based replication but obviously the choice is determined by your preferred operating platform. Oracle GoldenGate is another possibility for data replication if that is your preferred database.
Do you also have multiple client managers? Are they in each of the datacenters? If so, are both client managers active or are you only activating the client manager that is on the same datacenter as the live master database?
If you have both CM active on two separate DCs, do you see any noticeable difference on performance between the CM that is on the same DC as the master database compared to the one that is on the other DC?
We have the same implementation as you on our 5.3 PRD, but because we are only in 6.1 on our DEV - which doesn't span across DCs, we don't have a good idea on how TES 6.1 or 6.2 CM performance will be in the multi DC setup.
We are also still in the process of rolling out T6 so it sounds like we're in a similar position.
So far we've found that the java client suits our users better than the CM/web interface. Obviously this means we have to spec up the desktops but the performance for us is better using the java client.
We do have dual CMs but at the moment they will probably end us as backup for the java client.
We were impressed with the java client's speed performance too, but stability wise since it was just released we felt more comfortable waiting for it to mature more. So will most likely proceed with the client managers only in 6.2.1.
It is worrisome to not know how performance will be with multi DC with the CMs. The java client would have been perfect since it greatly simplified the backend architecture - just like going back to 5.3.
Last question, are you using VMs? Did you make your resources (CPU/RAM) dedicated like Cisco recommended?
We are using VMs currently although our original T6 dev work was done on discreet hardware. So far I haven't seen any appreciable difference. Our upgrade includes a platform migration so we're coming off SPARC/solaris and going to X64/linux.
The new stuff is certainly quicker but then SPARCs always were a bit slow.