i have been using the following suggestions
raid 01 for mdf ,ndf
raid 1 (duplexed) for ldf
create the database (system tables mdf)
then create a seperate file ndf for the data portions on your datastore.
place this ndf in a seperate filegroup and make the new filegroup the default filegroup.
this causes all objects to be created in the new filegroup, away from the io that reading and writing to the database catalog will create as users try to gain access to the database and it's objects.
keep the log on a seperate mirrored (duplexed) drive to avoid the direct disk contention that anything on the disk with the log will generate.
try not to use raid 5 due to the excessive io that comes with each disk write.
with this said, understand that the bean counters wont allocate funds for you so you may end up installing on an thomas edison Phonograph Cylinder
whew all that work for that obscure reference