On Tue, 2007-08-14 at 14:12 -0700, Carl Lowenstein wrote: > On 8/14/07, kelsey hudson <[EMAIL PROTECTED]> wrote: > > Ralph Shumaker wrote: > > > > > Would a SCSI transfer be quicker between two disks on one controller or > > > on separate controllers? If memory serves me, I think I remember > > > something about SCSI being able to simultaneously copy one disk to six > > > others all at the same time while barely touching the CPU or system > > > memory. Is this even close? (Memory may be foggy, and I may have > > > misunderstood it to begin with.) > > > > Faster? Maybe. SCSI *can* busmaster, true. And, the controller can be > > programmed to do what you say. > > > > You know, this is true in principle. But software support to send the > appropriate commands to the SCSI devices is scarce to non-existent. >
It's unfortunate that the industry mangled what SCSI was designed for. It was not designed just for connecting scanners and hard drives to a computer. It was designed to connect *any* device with a SCSI interface to a common system bus. The SCSI bus was designed to allow multiple hosts and every device could become a host or a slave. In 1988 I designed a video system (while I still worked at NCR, the folks that brought us SCSI, but it was not done for NCR) that had a SCSI interface on board so that it could load video data directly from disk, bypassing the host computer. Such a system would have a minimum of two hosts - the host computer and the video system - and multiple slaves. To this day I'd still like to see video systems with smart peripheral interfaces on them. It would greatly speed up system performance. As Carl stated, software to do such things is scarce to non-existent, thanks to the limited adoption of SCSI by the industry. PGA -- Paul G. Allen BSIT/SE Owner/Sr. Engineer Random Logic Consulting www.randomlogic.com -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
