I assumed the first 3950 would be host controller 0 and 1
and the second 2 for channel A and 3 for B.
With the second controller installed I loose the cdrom and tape
(controller 1, Channel B) and I still don't see the hd with mkdev hd
using 2 as the host controller.
Is my assumption wrong re the numbering scheme? Is what I'm trying to do
possible?
Any help greatly appreciated. BTW, I'm using the blad driver.
Heinz
on site in Bakersfield.... many miles from home :-(
What are you trying to accomplish here?
Only 2 HD's on the wide channel should not result in any noticeable
speed decrease. Unless you're trying for redundancy on mirrored drives?
If you're trying not slowdown during backups, that's pretty much going
to happen anyway unless you have one of the rare tape drives with SCSI2
wide cables. I've never seen one with UW.
Even then, there's no way to physically move the tape fast enough to
match the throughput from a fast UW drive.
Sorry, I've never installed two 3950U2's in one SCO box, what does
hwconfig report for the adaptor ID's, and what does your BIOS say about
the cards?
--
---------------------------------------------
Pat Welch, UBB Computer Services
SCO Authorized Reseller
Unix/Hardware/BB Sales & Support
Year 2000: Consulting and Repairs
*** 96 Days left 'til 1/1/2000 **
(209) 745-1401 Fax: (209) 745-5640
Nationwide pager: (800) 608-7122
E-mail: pat...@inreach.com
----------------------------------------------
>I'm trying to run 2 controllers so that I don't bring the first's
>Channel B down to narrow scsi speed (tape and cdrom) for the second
>HD.
You are making assumptions of how SCSI works - perpetated by
articles in the popular PC press and even in documentation in at
least on well known certification manual.
The bus slows down to the speed of the slowest device ONLY when
that device is communicating with another device on the same bus.
There can only be two devices on the bus at one time. The devices
negotiate their highest common speed. For example, if you had
a SCSI I then the highest bus-speed would be 5MB/sec (and so many
of the earlier devices couldn't even touch that rate in data
transfer). Connect that to one of the newest Ultra drives
with a 40MB/sec bus speed.
The data transfer will occur at the 5MB/sec speed.
When the slow device is finished, then perhaps another 40MB
device will want to exchange data with the one from above.
At this point bus is running at 40MB.
IOW - the bus does not slow down to the slowest device on the bus
EXCEPT when that device is one of two devices communicating. The
bus speed changes depending on device connection.
If you put your slow device on another controller and want to
send data to the controller with the slower device, then you are
still limited by the slower device as to the rate of speed of the
data transfer - effectively slowing the output - not the bus speed
- of the device on the first controller.
In smaller systems you should not see any performance difference.
You increase the cost with no performance gain. In larger systems
if you have many slower devices you could put in a controller for
the lower speed devices and reserve the ultra style controller
for several HD's. For small systems, two or three HD's, that's
overkill IMO.
--
Bill Vermillion bv @ wjv.com
You are absolutely right in the assumption I was making, i.e. that the
device merely on the bus would cause the slowdown.
Thanks for the explanation. You have saved me grief and my customer
money.
Thanks again
Heinz
Bill Vermillion wrote:
>
> In article <37EFD39C...@bytedesigns.com>, Heinz Wittenbecher
> (at Pioneer) <he...@bytedesigns.com> wrote:
>
> >I'm trying to run 2 controllers so that I don't bring the first's
> >Channel B down to narrow scsi speed (tape and cdrom) for the second
> >HD.
>
> You are making assumptions of how SCSI works - perpetated by
> articles in the popular PC press and even in documentation in at
> least on well known certification manual.
>
> The bus slows down to the speed of the slowest device ONLY when
> that device is communicating with another device on the same bus.
>
<snip>