what do you want to hear about?
there are some papers available at OTN
Some basic points?
* yes, it works
* no you cannot mix different OS/platforms
* yes, it scales, IF your application will scale on SMP-machines
* yes, you need a single storage accessable for all Instances of the same DB
* yes, you need a OS-clusterware
* no, TAF (transparent failover) is not transparent in all cases.
do you have detailled questions so they might be answered.