Date: Sun, 21 Feb 1999 00:29:53 -0700 (MST) From: "Kenneth D. Merry" <ken@plutotech.com> To: dennis.glatting@plaintalk.bellevue.wa.us Cc: freebsd-current@FreeBSD.ORG Subject: Re: Slow seq. write on Seagate ST36530N Message-ID: <199902210729.AAA17355@panzer.plutotech.com> In-Reply-To: <199902210212.SAA07089@imo.plaintalk.bellevue.wa.us> from Dennis Glatting at "Feb 20, 1999 6:12:14 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
Dennis Glatting wrote... > > > > Just to throw my voice in, I get crappy SCSI performance on my > > > three SCSI card Dell. I don't understand why. > > > > Umm, you're going to have to be a little more specific than that. > > > > You do have one of those slow Western Digital drives, which > > could cause some problems. > > > > The WD drive is on its own controller. The ST410800W is on its own > controller. The two ST43400Ns are together on the third > controller. Well, by "cause some problems", I meant that those disks are known to have bad performance when tagged queueing is turned on. That's why we have a quirk entry in the transport layer to turn off tagged queueing for Western Digital Enterprise disks. > On the ST410800W I am getting, according to iozone: FWIW, several people have reported bad performance with that drive when tagged queueing is enabled. Their firmware revisions, however, were 71xx, not 45xx. Apparantly drives with firmware that starts with 00 work okay. da1 at ahc1 bus 0 target 0 lun 0 da1: <SEAGATE ST410800W 4508> Fixed Direct Access SCSI-2 device da1: 20.0MB/s transfers (10.0MHz, offset 8, 16bit), Tagged Queueing Enabled da1: 8669MB (17755614 512 byte sectors: 64H 32S/T 8669C) I haven't seen one of those Elite 9's with 45xx firmware before, so I suppose it's yet to be determined what the problem is. > > btw# iozone 1024 8192 /ExternalDisk/junk > > .... > > Writing the 1024 Megabyte file, '/ExternalDisk/junk'...299.609375 seconds > > Reading the file...178.609375 seconds > > > > IOZONE performance measurements: > > 3583805 bytes/second for writing the file > > 6011676 bytes/second for reading the file > > > > Since this is a 16 bit device with 10 MHz transfer rate I expect > better, or should I not? According to the disk specs the average > internal transfer rate is 7.2 MB/sec, 20 MB/sec external > burst. I would expect better, at least for write performance. > The following is an examination of the device's mode pages. I'm > currently looking at each variable and their meaning, though I > am not a SCSI guru. Well, there are a few things you can try: > > ==> 8 > > > > IC: 0 > > ABPF: 0 > > CAP: 1 > > DISC: 1 > > SIZE: 0 > > WCE: 1 Try disabling write caching, by setting the WCE bit in mode page 8 to 0. > > MF: 0 > > RCD: 0 > > Demand Retention Priority: 0 > > Write Retention Priority: 0 > > Disable Pre-fetch Transfer Length: 65535 > > Minimum Pre-fetch: 0 > > Maximum Pre-fetch: 29 > > Maximum Pre-fetch Ceiling: 477 > > > > > > > > > > ==> 10 > > > > RLEC: 0 > > Queue Algorithm Modifier: 1 > > QErr: 0 > > DQue: 0 If that doesn't work, try disabling tagged queueing by setting the DQue bit in mode page 10 to 1. I believe we only read those settings at probe time, so you'll probably have to reboot if you set that value. > > EECA: 0 > > RAENP: 0 > > UAAENP: 0 > > EAENP: 0 > > Ready AEN Holdoff Period: 0 If it turns out that tagged queueing is the problem, we may need to do some more diagnostics. 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?199902210729.AAA17355>