Date: Thu, 6 Apr 1995 21:13:50 -0700 (PDT) From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> To: terry@cs.weber.edu (Terry Lambert) Cc: freebsd-hackers@FreeBSD.org Subject: Re: large filesystems/multiple disks [RAID] Message-ID: <199504070413.VAA06032@gndrsh.aac.dev.com> In-Reply-To: <9504062314.AA16204@cs.weber.edu> from "Terry Lambert" at Apr 6, 95 05:14:15 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > > > > > > ftp://csvax.cs.caltech.edu/pub/stripe.tar > > > > > > which might be somewhat helpful - a rather dated (4.2?) attempt at > > > a simplistic striping device driver. > > > > Any one want to buy my Quantum Empire 2100 for $800, it's less than 120 days > > old, as I am now going to sell it off and go buy 4 very fast 500MB > > drive and make a stripe driver out of this code :-). > > This might be going overboard. Is there any reason you can't divide it > up into four logical drives and stripe it that way? Save you eating > some $ on your disk drives (the better to spend on coffee, I suppose). But doing that won't make things go any faster, and the whole reason I want to do this is 5MB/sec is not fast enough for my tastes. Also I have a real system running on that 2 G drive, if I replace it with 4 drives I could put my system on 2 drives pretty easily and have 2 whole drives to play with. If I try to slice up my 2 G drive I run the chance of creaming my running system :-(, and I won't be able to measure concurrency :-(. > > As long as you are careful about your placement of your synchronization > primitives, it shouldn't be an issue that you're really only using one > drive. I will test that, there are even some comments in the README about making it SMP safe :-). > > How low level is the code? I haven't had a chance to look at it yet > myself... if it's SCSI command level, then forget I said anything and > go for the multiple drives; otherwise, it'd be nice if it were general > enough to use on anything for which a device could be obtained (ESDI > or MFM layered above the media correction, etc.). Pretending to have > multiple drives on a single drive would be "good practice" for getting > the code to be as general as possible... It is general, it runs above the bdev/cdev layer. It could care less about the device itself. Looks very similiar to the way you set up stripes on an Auspex. >From the README: | To create a striped partition, you need to supply an interleave modulus, | which must be a multiple of MAXBSIZE (specified in DEV_BSIZE units). | It should evenly divide the real disk partitions that are being | striped, or at least all but one of them, if you want to be able | to use all of the space. The real disk partitions to be striped | are specified as two numbers: a major device number and a minor | device number, separated by spaces. In my /etc/rc, I have: | | /etc/ilvconfig /dev/rilv128 960 0 29 0 45 >/dev/console | | to stripe together these partitions: | brw------- 1 root 0, 29 Apr 21 1987 /dev/hp3f | brw------- 1 root 0, 45 Dec 22 00:39 /dev/hp5f | | This has to go before the invocation of fsck, so it's about the | third line in the file. -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Custom computers for FreeBSD
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199504070413.VAA06032>