A Volume Manager Fundamentals presentation by Veritas states:

"Do not put log subdisks on a heavily used disk."

Does anyone know if this relates to an i/o
performance impact? or is there something else?

I have seen a scenario of a stripe which contains 4 mirrored
sub-volumes and struggles with 50+ %busy on i/o (and I see
corresponding high io waits on the CPU)a lot of the time.
The entire volume consists of 8 disks - 7 of which are used
for data only - 1 of which is used for the DRL subdisk for
each of the 4 sub-volumes and also for data.
So the same disk is sliced 5 ways - 4 slices for DRL and 1
slice for data. The volume is used for Oracle redo on a
very busy mail server. What I'm thinking is that this type
of layout is going to cause a bottleneck with not only data
being written to this one disk, but also the DRL for all the
subvolumes - any thoughts?

Also, if the disk containing all the DRL for each subvolume
and the data for one side of the mirror of one subvolume were
to fail......
won't this make having DRL in the first place completely pointless?