I don't get the power saving thing though - that sounded very snake oil like. I mean, if your system isn't compromised, what CPU operations is it reducing exactly?
There is a bit in the linked PDF which says...
"Abatis Hard Disk Firewall, was also tested using the same standardised environment and shown to block applications and background processes from executing; saving energy from a baseline configuration."
What they seem to have done in the test is taken a standard system and measured the power consumption. They've then tested that baseline with one of 3 3rd-party AV products and recorded the power consumption go up. They've then installed tested it with their kernel module that blocks I/O and unsurprisingly noticed that a system which isn't using the disks uses less power.
It also says...
"Between best case, HDF and worst case, AV Product 2 there is a potential annual cost saving in excess of Â£12 at server level, this scaling up to Â£125,000 in a data centre with 10,000 servers."
I would have thought that if you had 10,000 servers and wanted to avoid power I/O costs you wouldn't have specced them with physical storage in the first place and would be network booting them instead.