Date: Mon, 8 Sep 1997 10:00:02 -0700 (PDT) From: lamaster@george.arc.nasa.gov To: current@FreeBSD.ORG Subject: Re: lousy disk perf. under cpu load (was IDE vs SCSI) Message-ID: <199709081700.KAA15180@george.arc.nasa.gov>
next in thread | raw e-mail | index | archive | help
Amancio Hasty <hasty@rah.star-gate.com> writes: |> I have to say the same thing over here. IDE disk drives over here |> tend to go belly up and my scsi drives tend to work a lot longer. |> |> >From The Desk Of "Michael L. VanLoon -- HeadCandy.com" : |> > The standard qualifiers apply. This is one guys' experience, on a I think the choir is in complete agreement on all the major points discussed: SCSI is usually higher quality; SCSI usually has higher performance; EIDE is usually cheaper; when the performance and quality are about the same, usually the cost is about the same, too; SCSI requires an extra controller, but, cheap NCR/Symbios controllers (e.g. $80) make this mostly moot; most of us use IDE/EIDE at least sometimes on legacy hardware, and even, for legacy-compatibility reasons, actually buy EIDE occasionally, and so IDE/EIDE support is important anyway; expandability, flexibility, tagged queueing, IRQ conservation, etc. etc. all favor SCSI; cost, Wintel compatibility favor EIDE. And now for the obligatory confessional: Speaking personally, I have always ended up regretting every dime I have spent on (E)IDE, as opposed to SCSI, as rational as it seemed at the moment I did it for perfectly good at the time legacy-compatibility reasons. So, if I were recommending to someone, I would recommend SCSI. (I have also regretted every ISA card, and wish there were a "standard" PCI sound card.) Of course, IMHO, personal opinion, YMMV. Now then: There is one issue I would like to return to, and that is the question originally posed: lousy disk performance under *some* circumstances. (Which circumstances?) "Mats Lofkvist" <mal@algonet.se> originally asked the question in this thread, but, I have observed a problem, too, and, the interesting thing is that not everybody seems to experience the problem in the same way. In my case, I am interested in improving "make world" time. I'm not in a position to reconfigure at the moment, but, eventually, I will try switching Buslogic and NCR controllers and see if the NCR driver's tagged queueing speeds things up. I will also add a separate /usr/obj partition/filesystem which is expendable (mounted -o asynch,noatime), which reportedly also speeds things up tremendously. But, I'm also wondering if there are any kernel config options or startup config options wrt filesystems, scheduler, or anything else that might explain the large differences between what I and some others experience and what some of you wizards out there are getting (e.g. 1 to 1.5 hr elapsed time for "make world")? Or is the /usr/obj filesystem the only big optimization available? -Hugh LaMaster Corrections welcome. Hugh LaMaster, M/S 258-5, ASCII Email: hlamaster@mail.arc.nasa.gov NASA Ames Research Center Or: lamaster@nas.nasa.gov Moffett Field, CA 94035-1000 No Junkmail: USC 18 section 2701 Phone: 415/604-1056 Disclaimer: Unofficial, personal *opinion*.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199709081700.KAA15180>