[email protected] (David Boyes) writes:
> What about things like RSCS v1? I think it could run on bare metal in
> its early days (morphed later into becoming GCS-dependent in RSCS v2).

Original CPREMOTE ... straight point-to-point might run on bare metal
... but vnet/rscs used for the internal network had a least spool block
diagnose ... used VM370 spool file for intermediate storage
(store&forward) and transfer on local machine.

In the 80s, the CSC co-worker responsible for VNET/RSCS had left IBM and
was working on advanced real-time system projects and thot that parts of
the dominant industry real-time operating system looked familiar. He
compared the real-time system monitor with RSCS monitor and it turned
out to be almost line-by-line translation of 370 assembler to "C"
... including all the original comments.

The original announcement in 1976 was joint with JES2 NJI. The problem
was that mainstream IBM was still having problems adapting to unbundling
announcement and charging for application software. Process was forecast
at high, middle, low price and price*numbers had to cover deevelopment
and support of the product. There was no forecast for NJI that covered
cost. On the other hand VNET/RSCS could get away with $30/month. They
fiddled the announcement as combined NJI & VNET at $600/month ... where
the VNET revenue covered the NJI shortfall.

VNET/RSCS had correctly layered ... but NJI intermixed network fields
with job control fields. The result was that VNET/RSCS easily supported
both NJI drivers and native drivers (that had much higher throughput
than NJI drivers). The intermixing of JES2/NJI fields that traffic
between different releases of JES2 could crash the host MVS. The
internal network was larger than arpanet/internet from just about the
beginning until sometime mid-80s ... mostly RSCS/VNET ... with
sprinkling of MVS/JES2. NJI implementation came from HASP (the source
carried "TUCC" in cols 68-71) and used spare entries in the 255 psuedo
device table to define network nodes ... maybe 150-200 max. nodes ...
and would trash traffic if either origin or destination nodes weren't
locally defined. By the time NJI/VNET shipped to customers, the internal
network had passed 255 nodes ... so there issue that JES2 would discard
traffic (when it wasn't crashing MVS).

Internally, MVS/JES2 were restricted to edge nodes (to minimize trashing
traffic) fronted by RSCS/VNET with special NJI driver that could rewrite
NJI header information to match the directly connected JES2/NJI release.

At the time arpanet/internet had great switch-over to internetworking
protocol on 1Jan1983, it had around 100 IMP network nodes with around
255 connected hosts ... while the internal network was rapidly
approaching 1000 (which it passed couple months later).
http://www.garlic.com/~lynn/subtopic.html#internalnet
and
http://www.garlic.com/~lynn/submain.html#hasp

The corporate sponsored university network used similar RSCS/VNET
technology as the internal network
https://en.wikipedia.org/wiki/BITNET
and
http://www.garlic.com/~lynn/subnetwork.html#bitnet

but were increasingly being forced to only ship NJI drivers (even tho
native drivers had much better performance ... and then later were later
being forced to SNA/VTAM (GCS).

Inside IBM I had HSDT project
http://www.garlic.com/~lynn/subnetwork.html#hsdt

in early 80s, starting to support full T1 and faster speed links. This
caused lots of problems in the communication group since 37x5 boxes only
supported 56kbits/sec. I also had problems with the RSCS 4k block spool
file diagnose interface ... since it was syncronous (RSCS didn't execute
while diagnose was in operation) ... and with competition from other
users of spool file ... RCSC might only get 3-8 4k blocks/sec
(12kbytes-30kbytes) ... and I needed aggregate of couple mbytes/sec.  I
needed asyncronous, non-blocking diagnose interface, and I needed vm370
spool file to do contiguous allocation, read-ahead, write-behind,
multi-block, even full track transfers, old mentioning SFS (spool file
system rewrite)
http://www.garlic.com/~lynn/2005s.html#28 MVCIN instruction

After the mid-80s, I was trying to get the internal network backbone
enhanced to support T1 & faster-speed links as well as enhanced spool
file operation. The communication group then got the backbone meetings
restricted to management only ... as part of forcing move to SNA/VTAM
... not letting technical people confuse the issues. recent discussion
http://www.garlic.com/~lynn/2017d.html#72 more IBM online systems
old email
http://www.garlic.com/~lynn/2011e.html#email870204
and then
http://www.garlic.com/~lynn/2006x.html#email870302
http://www.garlic.com/~lynn/2011.html#email870306

-- 
virtualization experience starting Jan1968, online at home since Mar1970

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to