[TUHS] networking on unix before uucp

[Lyndon Nerenberg]

On Aug 25, 2014, at 12:20 PM, arnold at skeeve.com wrote:

> Now add a few blits, where each physical terminal is doing the load of
> 4-6 virtual ones (using pseudo-ttys) - a shell on each pty and commands
> running on each pty.  Bingo! The load average shoots way up.
> 
> So, the blit itself wasn't at fault.  All the people taking advantage
> of what it could let them do, was.  At some point they wanted the lab
> staff to stop using them because of this...

Not just the blit.  When the Telebit Trailblazer modems came out I was one of the first Canadian resellers.  Demoing those beasts was a tricky proposition.  I would haul one out to a customer site, wire it up to a serial port, then dial up to our office system and let the beast loose.

They made a great torture test for mid/late-80s tty drivers (and RS-232 hardware interfaces).  Simply 'cu'ing to the office over the modem link, then 'cat'ing a 100 K text file over the link would reliably take a 3B2 to its knees.  After the first couple of demo's – which invariably brought the local staff out of their offices to ask of the machine had crashed – I learned to schedule these over the lunch hour, or after office hours :-)

It was quite astounding to see the wide range of performance impacts this had on various systems.  3B* systems would tip over and die, except for the (built by Convergent Tech) 3B1.

Suns with VME-based serial cards performed quite well.  If anyone remembers 'ncc' on the UUCP network, it was a 3/280 with the 16(+?)-port  VME serial expansion board, a Telebit rackmount chassis populated with eight modem cards, and another eight Convergent NGEN workstations cabled up to do file transfer.  Even with all the Telebits running UUCP flat out at full speed, you couldn't tell it from the interactive response time on the terminals.

I also used to own an NBI UNIX system.  This was an interesting little beast.  It was a QBUS machine in a deskside tower case that really wanted to be a VAX, but it had on a 68010 processor.  It ran a very generic port of 4.2BSD (complete with the Pascal interpreter/compiler!).  The machine came with the QBUS equivalent of the VAX DH11 serial board, but there was a bug in NBI's driver for the board.  I lifted the DH11 driver from the UCB 4.2 tape, changed the name of one struct, compiled and linked a new kernel, and suddenly it, too, was capable of sustaining several high speed Telebit UUCP links.  (That machine was called 'canada' for anyone still keeping track.)

Fun times :-)

-- {alberta,pyramid,uwvax}!ncc!lyndon

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20140825/db604cc2/attachment.sig>