Dear Miguel
The answers to your questions lie in IBM i Performance Data Investigator (PDI) charts. You should start with the chart named "Wait Overview" from both vdisk host and client LPARs to see what kind of wait time is (or are) dominant and then address it (them).
here is an example :
This Wait Overview chart from vdisk client LPAR shows that during the day time (in which the customer wants better performance), Disk Page Fault time appears most frequently and with the most quantity with CPU Queuing time the next frequent and quantity.
High and/or frequent Disk Page Fault time means disk response time of the LPAR may not be consistently good. So we look at the chart named "Disk Throughput Overview for Disk Pools".
Now, this customer's server has SSDs and overall disk response time of SSD that is higher than 2-3 msec. consistently is considered bad as seen above.
Next, we look at the same chart from the vdisk host LPAR and see that its overall disk response time NEVER goes beyond 0.6 msec. which is consistently good.
So, it is clear in this example that there is quite a bad disk response time overhead from which vdisk client LPAR suffers. In this case, I suggest the customer open the service call to IBM to investigate for a remedy.
As for the CPU Queuing that is the 2nd dominant wait from the overview chart above, it is more straightforward to address: either allocate more CPU power or control or try to streamline the day-time workload.
If your Wait Overview charts have different wait signature from this example and you want to see what can be done. post the charts here and I may help analyze them for you.