We're using Timefinder and ADSM to backup about 800GB of Oracle Raw
volumes, using adsmpipe from a dedicated "twintailed" node to throw the
data to ADSM once it's been synchronized and the BCVs split.  The
biggest advantage to the approach is that it allows you to minimize the
impact of backup activities on the database.  You can run a continuous
hot backup with your tables in backup mode for only a minute or so per
day, greatly improving performance and decreasing the amount of archive
logs you cut.

The biggest problems we've had are related to our decision to not create
a complete BCV set.  Our primary data sources are in striped VGs
containing 16-40GB of data.  In order to save some money we've created
pairs of BCV sets that allow us to mirror one volume group while backing
up another, sort of a double buffering strategy. I personally don't
recommend the approach (and this is just my opinion, guys), and would
strongly recommend a complete set of BCV disks instead.  If anyone's
interested in more info, send me some email at bill AT megsinet DOT net.