Date: Wed, 30 Mar 2005 14:02:22 -0800 From: John-Mark Gurney <gurney_j@resnet.uoregon.edu> To: Martin Nilsson <martin@gneto.com> Cc: freebsd-current@FreeBSD.org Subject: Re: ciss(4): speed degradation for Compaq Smart Array [edited] Message-ID: <20050330220221.GH37984@funkthat.com> In-Reply-To: <424B0614.5000902@gneto.com> References: <20050330191824.4c08acc6.aka@veco.ru> <20050330090813.B64732@carver.gumbysoft.com> <424B0614.5000902@gneto.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Martin Nilsson wrote this message on Wed, Mar 30, 2005 at 22:03 +0200: > Doug White wrote / skrev: > > > >I'd also suggest using a tool like iozone to run your tests instead of dd. > >Unless your workload consists of entirely sequential writes this perf test > >is worthless. > > What is wrong with using dd to measure max read/write bandwith of the > controller-disk system? It's not just measuring the bandwidth of the disk system, since it also must read from somewhere first.. So all of your numbers read x bytes from /dev/zero (or else where) and writes x bytes to disk, or read x bytes from disk and writes x bytes to /dev/zero... It is much more optimial to skip the read/write to /dev/zero, and programs like bonnie do this... (Though the cost of modern read/write from/to /dev/zero, it is still measurable.) -- John-Mark Gurney Voice: +1 415 225 5579 "All that I will do, has been done, All that I have, has not."
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050330220221.GH37984>