I've stumbled upon an issue with SAN storage, where we have virtual and physical GB. It is thereby possible to over allocate a higher number of virtual GB, than there physically exists.

Which number should be used in the mapping - The physical or virtual number?

This is known for virtual vs physical servers, where we know that the count of CPU or Memory for a virtual machine is actually vCPU and vRAM.

I think it's useful to look at this from the angle of "what insights is the user trying to gain from examining cost and utilisation of their storage environment?"

So:

"Have I got enough storage? Too much, too little, just right?"

"What is my Cost per GB against other storage options?"

There are a lot of potential questions - such as those addressing CIA, capacity and ability to expand and reduce storage, portability, performance and tiering, along with the inevitable questions about why Enterprise Storage is so expensive compared to a USB stick, but in general, they can be boiled down to the two above.

Therefore, in terms of storage virtualisation, I believe that we should be examining the total storage that can be accessed without compromising performance - in other words, the manufacturers guidelines for the amount by which a storage environment can be overprovisioned from its available physical capacity without compromising performance. Note that this may differ from what is actually implemented - but an excessively overprovisioned storage environment can lead to significant performance issues, and so we should take care to avoid portraying this as a good thing.

There is a side note here, which is that in an environment that mixes traditional storage with storage virtualisation devices, an additional model object should be considered to reflect this accurately, similar to the Physical Server - Hypervisor - Server triangle - but that falls into a different discussion.