Re: dm-crypt/LUKS performance
On Sun, 16 Nov 2014, David Fuchs wrote:
> I am setting up a system with an Intel octo-core Avoton, which has AES-NI
> support. After doing some crude benchmarking tests with dd, I am surprised
> about the huge performance penalty that full-disk encryption apparently has
> on read/write throughput.
You need to use a recent kernel that can run dm-crypt in parallel (and it
needs to be compiled with that option enabled as well. I don't know if
Debian's 3.16 is compiled like that). That information is missing from your
> The system will be used as a home file server, and the results with drive
> encryption are still acceptable - but I'm still curious if they are to be
> expected, or if there is an obvious culprit for the performance hit. Is it
> possible that I'm not using the hardware AES?
Check in /proc/crypto whether aes-asm is listed, and whether it has a higher
priority than aes-generic.
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot