Before 3.0.3 it would create one vmdk containing all volumes of a server. We always create seperate vmdks for the disks of a virtual machine. Before 3.0.3 we were forced to use
platespin because of this. Now vmware converter is also an option for us.
Frank Brix Pedersen
Personally I do use the option to create a seperate VMDK for each volume. I prefer this option as it makes life much easier if you need to add space to a volume at a later date.
One factor that should be taken into account is how many VMDK's you will have per LUN. This is also dependent on the amount of I/O generated by the VM's. If you have intensive I/O VM's keeping the number of VMDK's per LUN below 10 is a good idea. For high I/O I would stay below 16. As is the norm in Virtualisation balance is the key, if you have a few high I/O servers put these on their own LUNs then put lower I/O servers with them. This will balance it out.
Check out my blog at http://www.ridethevirt.blogspot.com
Prior to the latest version, P2V would attempt to create one VMDK per PHYSICAL DISK SET, not per server. (If you had say 2 disks mirrored and another 4 or 5 in a RAID 5) P2V would create a single VMDK for each disk set regardless to how many partitions there may be on them. Now P2V will let you a separate VMDK per each partition.
in the past, to split them up, we would run separate "passes" to create the separate VMDK's and then reassemble them afterwards. (What a pain!)
as stated before 3.0.3 it would create 1 vmdk and platespin powerconvert was a prefered option as it could split it out into additional disks, now with 3.0.3 now need to use powerconvert