Date: Sat, 10 Sep 2005 22:24:19 -0400 From: Nick Evans <nevans@syphen.net> To: "ke.han" <ke.han@redstarling.com> Cc: freebsd-geom <freebsd-geom@freebsd.org> Subject: Re: GELI hardware performance. Message-ID: <20050910222419.0e9dadaf@speedstar.syphen.net> In-Reply-To: <432380BA.209@redstarling.com> References: <20050910165926.40c5354d@speedstar.syphen.net> <432380BA.209@redstarling.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 11 Sep 2005 08:56:26 +0800 "ke.han" <ke.han@redstarling.com> wrote: > Thanks for the comparison. As a newcomer to both GELI and gbde, it > would be niced to know how these numbers compare to no encryption on > the same hardware. How hard would it be to run these performance > tests on the same hardware with no GELI? > It seems that for drives that put through over 100Mbs that your > results of around 14 to 27 Mbs (block output) are pretty low. Am I > readin the numbers wrong? comparing apples to oranges? > any help in my understanding is appreciated as I'm trying to decide > how to solve a new server setup. > thanks, ke han > No GELI, same disk: Per char output: 45967 K/sec, 82.2% CPU Block output: 56019 K/sec, 35.9% CPU Rewrite: 15833 K/sec, 10.5% CPU Per char input: 31412 K/sec, 60.8% CPU Block input: 40143 K/sec, 17.5% CPU Seeks/sec: 853.2, 5.2% CPU Looks like ~50-85% of the original performance when hardware crypto is used. Not bad considering.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050910222419.0e9dadaf>