Date: Sat, 20 Feb 1999 17:21:19 -0700 (MST) From: "Kenneth D. Merry" <ken@plutotech.com> To: mjacob@feral.com Cc: paulz@trantor.xs4all.nl, ken@plutotech.com, freebsd-current@FreeBSD.ORG Subject: Re: Slow seq. write on Seagate ST36530N Message-ID: <199902210021.RAA15371@panzer.plutotech.com> In-Reply-To: <Pine.LNX.4.04.9902201617050.31494-100000@feral-gw> from Matthew Jacob at "Feb 20, 1999 4:17:30 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
Matthew Jacob wrote... > > If you're feeling kind, could you try and see if you could get a > performance curve related to the number of outstanding tags allowed? I will send him some diffs for camcontrol in a separate piece of mail that will enable him to change the number of tags on the fly. > On Sun, 21 Feb 1999, Paul van der Zwan wrote: > > > > > > > > > This did not improve anything , but I think I have found the couse. > > > > In that modepage there is a DISC value which was 0 on the IBM and 1 on the > > > > Seagate. I remembered a ' Enable disconnect' option in the Adaptec 2940 bios, > > > > setting this to 'off' for both harddisks led to a huge performance increase on > > > > the Seagate. If I also enable Ultra mode iozone write goes from 1.5 MB/s > > > > to 12 MB/s ( a factor of 8 !!!). > > > > > > The 'DISC' bit in mode page 8 has nothing to do with disconnection. Here's > > > the description of it from the SCSI-3 spec: > > > > > > ================ > > > The discontinuity (DISC) bit, when one, requests that the device server > > > continue the pre-fetch across time discontinuities, such as across > > > cylinders (or tracks in an embedded servo device), up to the limits of the > > > buffer, or segment, space available for the pre-fetch. When zero, the DISC > > > requests that pre-fetches be truncated (or wrapped) at time discontinuities. > > > ================ > > > > I did not change it but it triggered me to have a look at the Adaptec BIOS > > settings. > > I have done some testing and reenabled disconnection, preformance dropped to > > less than 2MB/s writing. > > After searching the freebsd-scsi archives I found out it might be related to > > tagging, so I edited /usr/src/sys/cam/cam_xpt.c and added an entry to disallow > > tagging for this drive. Back to 12 MB/s. So it looks like there is at least an > > acceptable ( to me ) workaround. > > Maybe the CAM maintainer can put an entry in cam_xpt.c for this drive ??? > > dmgs shows is as : > > da1 at ahc0 bus 0 target 1 lun 0 > > da1: <SEAGATE ST36530N 1444> Fixed Direct Access SCSI-2 device > > da1: 20.000MB/s transfers (20.000MHz, offset 15) > > da1: 6208MB (12715920 512 byte sectors: 255H 63S/T 791C) As Matt said, it would be nice to figure out whether a reduced number of tags would do the trick. I'll send you some diffs for camcontrol that will enable you to tweak the number of available tags. (You'll have to remove the quirk entry you put in cam_xpt.c for that drive.) Ken -- Kenneth Merry ken@plutotech.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199902210021.RAA15371>