Date: Fri, 30 Jun 2000 04:09:31 -0700 (PDT) From: Jaye Mathisen <mrcpu@internetcds.com> To: "Koster, K.J." <K.J.Koster@kpn.com> Cc: hackers@FreeBSD.ORG Subject: RE: Funky scheduler stuff under heavy I/O. Message-ID: <Pine.BSF.4.21.0006300405160.18895-100000@schizo.cdsnet.net> In-Reply-To: <59063B5B4D98D311BC0D0001FA7E4522026D76D5@l04.research.kpn.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Well, I ran it by Soren, he didn't think so. And Originally when I noticed it, I thought it was vinum (although I'm not using vinum in this case), and mentioned it to greg, and he was leaning that way as well... I'm not sure I buy the controller argument anyway. For each pair of drives on each controller (Master/Slave), the 1st PCI Card/1st Controller Master == the 1st PCI card/1st Controller Slave which == the 2nd PCI card/1st Controller Master, etc... And the matching 2nd controllers stay almost locksteped with each other, but getting many fewer transactions than the 1st controllers on each card... So I don't see how it can be hardware. I don't believe there's any more cost to the IDE controller whether it's the 1st one on the card or not. The othe rthing is that Soren mentioned he was seeing the same thing on SCSI drives as well, which would eliminate the ATA code. Personally, I would've guessed that all the masters would've run faster than all the slaves, but it ain't so... On Fri, 30 Jun 2000, Koster, K.J. wrote: > > > > both the Master and Slave drives on the first channel of > > each controller are about 200MB's ahead of the master > > and slave drives on the second channel of each > > controller, and the gap is growing. > > > Umm. This sounds more like your controllers are doing this. You could stick > in another two IDE controllers and see if that cures the problem. You have > one of those 10 PCI bus boards, right? :-) > > Come to think of it, that would still not show if it's the hardware or > FreeBSD dropping the bucket. > > Just curious, but is using vinum actually improving performance? (Assuming > that's what you're trying to achieve) > > I've bonnied two Maxtor DiamondMax 15GB drives in my box, and I found that > when I stripe them, performance drops to 15MB/s, whereas a single Maxtor > will fling 25MB/s onto the platter. > > Kees Jan > > ================================================= > TV is the worst of both worlds. It's not as > good at words as radio is because the pictures > are a distraction which demand attention, and > it's not as good as cinema because the pictures > are not nearly as good. > Douglas Adams > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-hackers" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0006300405160.18895-100000>