From - Fri Apr 23 19:34:19 1999 Path: reader2.news.rcn.net!feed1.news.rcn.net!rcn!master.news.rcn.net!not-for-mail From: "Glenn C. Everhart" Newsgroups: comp.os.vms Subject: Re: Bad write performance under OpenVMS Date: Thu, 22 Apr 1999 20:13:52 -0500 Organization: GenCybEng Lines: 49 Message-ID: <371F8300.F5EF06E@gce.com> References: <199904220929.LAA04872@gate.fim.fgan.de> <623823988wnr@nugate.demon.co.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: KtynWfStAMqJ25Xb37K5XEfhNZ0HT+X059CH/IGIU+4= X-Complaints-To: abuse@rcn.com NNTP-Posting-Date: 23 Apr 1999 00:13:49 GMT X-Mailer: Mozilla 3.03Gold (X11; I; OpenVMS E7.1-1H1 DEC 3000 - M300LX) Xref: reader2.news.rcn.net comp.os.vms:228572 OK, at last I can comment on this having a SCSI standard in hand. The disk cache of a SCSI device in writeback mode means when you write to the disk, data goes to device cache, written to memory when the device feels like it, or when a SCSI synchronize cache command is given. VMS doesn't enable writeback mode, so also doesn't do synch cache. A cache for read data exists on the device; VMS can allow use of this and does. In addition there are lock or unlock cache commands. Lord knows if anyone implements them. Synchronize cache is by the way a separate scsi command (opcode 35 hex).(rest of 10 bytes can be zero.) The caching mode page is page 8. The format of the start of the modepage is 8 (page code) 0A hex (page length) WCE/MF/RCD in low 3 bits rest can be zeroes or whatnot. WCE = Write Cache Enable. 1 means writeback is ok; data need not be written to disk till the drive feels like it or gets a synch cache command. MF is mult factor, size of prefetch. Leave zero. RCD = Read Cache Disable. 0 means read data can come from cache or media. 1 means data must come from the media (so is slower). The VMS default is zero here. Set to 4 to enable write cache. This can be done with sys$etc:scsi_mode. DO IT ONLY for SCRATCH media. The filesystem you save will be your own. Dave Pickles wrote: > > In article: Jan > Vorbrueggen writes: > > > {...} In the > > usual case I'd think of as a write cache, the drive puts the data into the > > cache, tells the host it has finished, and only then (or concurrently) > > actually writes the data to the medium - this has different semantics in the > > case of failure. > > Pardon my ignorance, but isn't this what is meant by "write-back cache > enabled" which all RZ drives seem to do automatically under VMS nowadays. > > -- > Dave Pickles