Oh, thats interesting.
I can imagine that they would be very useful in azure to speed up VM-block storage communications. It possibly is secondary to reducing backbone bandwidth usage.
I assume that EC2 has lots of caching to make things viable (basic testing sorta backs that up)
They used to be quite popular in VFX land, as one could slap it infront of a filer and get a performance boost, without much work.
However once you move to a place where you can easily ans quickly move assets around, and have them on a larger amount of smaller servers (~150 tbs instead of a 1pb isilon cluster) then it becomes less relevant, especially for the price.
