terminals/term. servers

[Doug Gwyn]

In article <wZ_XRne00WB4I4M4kT at andrew.cmu.edu> sr16+ at andrew.cmu.edu (Seth Benjamin Rothenberg) writes:
>We were told that character input from hardwired terminals (as opposed to
>virtual terminals)  causes cpu interrupts.  Is this true?  Would this
>slow the machine down a great deal?

It depends on the serial line multiplexer and the operating system.
Most commonly encountered VAX serial line multiplexers are used by
most common UNIX implementations in a single interrupt per character
mode.  (There are exceptions, generally involving I/O processors.)

It's not a large problem for typical installations.

>Most of the systems I know of use telnet (or LAT) logins almost exclusively.
>Are hardwired terminals just as efficient?

The main advantage of Ethernet access is that, until the net gets
saturated, you can sustain higher throughput.  Most RS-232 terminals
cannot communicate faster than 19,200 bps.

>BTW, if we go to ethernet, should we leave the existing wiring
>in place in case of an "emergency"?

I wouldn't think it would help.  If your Ethernet stops working, it
would be less work to fix it than to convert from an Ethernet base
to an RS-232 one.