If you are writing a Block Stripesize, and whole stripes can be written at once, there will be no need to read the old data, because all the parity can be computed from the new-to-write data.

So, this may skew your formulas a little bit, and hopefully emphasizes the fact, that you need to be aware of the usecases of your storage subsystem. And you can tune important parameters of it (e.a. Stripesize) for bulk-data or small block access… But thats a different story, because you have to look into Filesystems and their behaviours …

Greetings