The management OS won't need much I/O once it's running, except when you start or stop VMs, take snapshots, etc. Give as much memory as you can to the guest OSes and they won't need as much I/O as they otherwise would.
Your major problem will be that there won't be much separation between the needs of one workload and the needs of another. Hyper-V will attempt to ensure that every guest OS gets fair time on the drive array. But that still won't ensure that performance is truly sufficient.
I'm not sure which recommendation you're siting. But note that the most common production Hyper-V configuration (with Windows Server 2008 R2) involves sharing a single LUN from some storage array with multiple hosts and putting all the VMs in the cluster on that shared LUN. So clearly separate spindles for each guest aren't required. What is required is sufficient IOPS for your workloads.