If you would have 4000 IOPS to leverage across related SQL Server disks, how would you do it?

We have rounded some numbers with specialized resources. But, it would be nice to collect some feedback from those who already fine tuned some of that.

Basically, given the following:

Disk 1 drive c: for the OS
Disk 1 drive d: for the database
Disk 2 drive e: for the log file
Disk 2 drive f: for the backup file

We all know the OS wouldn't require that much. So, around 500 IOPS would suffice. But, what about the remaining 3500. In your case, how did you use that 3500 among the other 3 drives?

We, of course, are looking for performance optimization in a virtual environment where the IOPS are controlled by IOMeter and/or at the Hyper-V level. So, as everything costs dollars, limitation applies and fine tuning like that is to be considered.