Date: Fri, 7 Apr 2000 08:56:13 -0700 From: Stephen Byan <Stephen.Byan@quantum.com> To: "'Kenneth D. Merry'" <ken@kdm.org>, Stephen Byan <Stephen.Byan@qntm.com> Cc: freebsd-scsi@FreeBSD.ORG Subject: RE: newfs on IBM disks slower than Seagate disks? Message-ID: <8133266FE373D11190CD00805FA768BF02EE9F66@shrcmsg1.tdh.qntm.com>
next in thread | raw e-mail | index | archive | help
Too bad. It'd be a useful thing for the I/O subsystem to know, both from a data integrity and from a performance viewpoint. In one of my former lives (at Hitachi), we hacked OSF/1's flavor of UFS to pass a "this is metadata" flag down to the disk device driver in the buf header. I arranged for the big honking mainframe disk controllers to put the unflagged writes in their copious volatile cache, and the flagged writes in their more-limited non-volatile cache. Regards, -Steve -----Original Message----- From: Kenneth D. Merry [mailto:ken@kdm.org] Sent: Friday, April 07, 2000 11:30 AM To: Stephen Byan Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: newfs on IBM disks slower than Seagate disks? On Fri, Apr 07, 2000 at 06:39:01 -0700, Stephen Byan wrote: > Does the FreeBSD SCSI subsystem set the FUA bit in the CDB for UFS metadata > writes? If so, then data integrity with WCE=1 is probably no worse than for > WCE=0, since the filesystem is caching non-metadata writes anyway. > > If UFS and CAM haven't made arrangements to hint which disk writes are > precious, then I think you're best off setting WCE=0, unless your system and > your disks are on a UPS. CAM doesn't set the FUA bit on metadata writes because it doesn't currently have a way to distinguish between metadata and normal data. Ken -- Kenneth Merry ken@kdm.org To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?8133266FE373D11190CD00805FA768BF02EE9F66>