IBMs recommendation for VDISK Size is to make them as large as you need them.
There is no benefit in splitting VDISKs in a custom size just to have a “loadbalance” over many vpaths for a LV.
You will not get any speed increase of many small VDISK, as the SVC behind this will handle all the requests anyway, and the maximum queue_lenght on the FCS adapters is by default 1xxx. If you reach this limit, it is not important if the requests come from many VDISKs or just from one.
Another note about SVC is that in the design phase it is quite important to choice the correct extend size for the mdisk-group. IBM recommends an extend size by atleast 64MB. Normally you should try and get an extend size by 128MB which will allow you to add up to 512TB of Space managed by the SVC Cluster.
Generally, and this is what I can confirm, is the usage of IBM SVC a performance improvement that you will notice if you hit the cache of the SVC.
The 8GB cache of the SVC is managed quite well, and you will notice jumps in the speed to atleast 100mb/s read/write. That was measured on a DS4700 System with 16x500GB SATA HDDs.