Date: Tue, 21 May 1996 07:40:51 -0400 (EDT) From: Peter Dufault <dufault@hda> To: salyzyn@inet.dpt.com (Salyzyn) Cc: scsi@freebsd.org Subject: Re: Re[2]: Who wants a DPT SCSI controller driver? Message-ID: <199605211140.HAA19424@hda> In-Reply-To: <9604208326.AA832625682@inet.dpt.com> from "Salyzyn" at May 20, 96 01:53:49 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > When we Stripe across controllers, it is done from within the > driver. It would be far better for the SCSI system writers to > place striping across controllers within that layer (as is > done under Linux). Can you clarify this part, since I don't understand what design is far better than ccd unless it is addressing a different problem and doesn't affect ccd. It makes sense to support striping against arbitrary block devices, as is done in ccd. This should not percolate down into individual controller devices (aside from possibly some get-capability types of functions with good defaults) since you'd like to just take an existing driver and plug it into the ccd system without any changes. It makes sense to stripe directly across a device that has special hooks for striping with direct driver support, bypassing ccd and using it as the designer intended. Is this what you mean? This isn't an argument against ccd as much as a statement that to use your controller technology properly you need a proper driver. (You could also mean that we need a layer of RAID control with a standard interface that is layered on top of a driver, and that driver could be either a ccd derivative or a board with RAID support, but I don't think you mean that). -- Peter Dufault Real-Time Machine Control and Simulation HD Associates, Inc. Voice: 508 433 6936 dufault@hda.com Fax: 508 433 5267
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199605211140.HAA19424>