On Wed, 15 May 2019 14:59:00 +0200, R.S. <r.skoru...@bremultibank.com.pl> wrote:
>In the old days there was a Storage Control Unit, i.e. 3830 and disk
>controller within disk cabinet, i.e. 3350-A2
>
>So, we have CPC-cable1-3830-cable2-3350A2controller-internal_cable3-disk.
>
>I'm trying to understand separation of duties between 3830 and 3350A2
>controller.
>What was defined as CU - it was 3830 or controller within 3350 cabinet?
>Which cable was a channel (Bus&Tag)? I guess it is "cable1" connecting
>CPC and 3830.
>
>Not to mention that some old reference manual's diagram shows yet
>another box between CPC and 3830 SCU.
Depending on the year, you might find (rusty memory):
host
- channel 0
- channel 1
- switch (2814/2914)
- channel 1
- Control unit 0 (bus & tag from channel/switch)
- device 0 ("string header") (A-Unit)
- device 1 (B unit)
- device 2 (B unit)
- device 3 (B unit)
- Control unit 1 (bus & tag) from CU 0
- repeat
- channel 2
- repeat
The A units handled the fan-out (signal and power) to dependent devices (B
units) in the string and had the logic to talk to the channel. The B units
were just dumb slaves to the A unit. The A units could only talk to B units
of the same type since all the power and signaling was custom. The interface
between the CU and the A unit was generally such that a CU could handle strings
of newer and older device types. The number of A units required, the number
of I/O devices included in an A unit, and the way B units were attached was
generally model specific, so you would see variations on the above. (Don't
confuse with more modern UNIT=3390B to indicate a PAV to MVS!)
It was sheer size of the componentry that drove this design. I think the 3990
was the last stand-alone disk controller. With the arrival of 2105s, the CU
was inside the same cabinet as the drives and Logical CUs (LCUs) were born.
One big black box (literally). Adding additional cabinets no longer affected
the I/O configuration - just capacity.
We still have switches, of course, but they're no longer pull-turn-push. :-)
Alan Altmark
IBM Systems Lab Services
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
