I've used the rule in the past of logs get mirrored and data gets RAID 5. It seems to be a losing battle against the network guys who like to buy TB size drives and treat sqlserver as if it worked like file systems. I ask for many separate medium size physical drives and they usually just stick database files on the NAS.

At a minimum separating clustered/heap data from nonclustered indexes, log, and tempdb will give pretty good performance. We spend time optimizing only to have the disk access get single-threaded in the end by lack of understanding with network guys.