386/ix uucp throughput still stinks

John G. De Armond jgd at rsiatl.UUCP
Mon Feb 19 09:29:17 AEST 1990

johnl at esegue.segue.boston.ma.us (John R. Levine) writes:

>It's widely known that 386/ix (2.0.2, new X5 driver in case you care) async
>throughput stinks, particularly if your async port has an unbuffered 8250
>UART.  So I got a 16550A, unsoldered the 8250 from my internal Telebit
>modem, soldered in a socket, and stuck in the 16550A.  I can send stuff to
>other sites at 1200-1400 cps, which is what I hoped for.  Inbound news is
>still a lousy 500 cps.  What gives?  The modem is set up as ttyd1, with tty00
>being a mouse that I don't use much.  That wouldn't be a problem, would it?


You might want to rig up some send and receive data indicators (LEDs or
whatever) and watch the data as you receive your feed.  I've noted on my
external telebits that few other sites, especially larger machines,
can keep the pipe full at 19200 bps.  Especially if your feed is working
from a modem server, your feed rate is pretty typical.  Here at RSI,
we get our primary uucp and News feed from emory which feeds its modems
over an ethernet link.  We generally average <800 cps.  the problem is
easy to observe with data lights.  The feed is bursty and rather regular
which probably means small packets on a busy ethernet.

Remember too, that if you have spoofing turned on, the modem handles
the ACK traffic and is the primary throttle to receiving speed.  If your
machine cannot for some reason, handle the incomming rate, you should see
alarms in the uucico debug output.  This symptom is very easy to spot with
data lights.

I will note that on this machine, a 20 mhz non-cached 386 clone, the X5
drivers would not reliably handle 19.2kb if any other jobs were running.
Cron could kick something off and immediately characters would be dropped.
The alarms in the debug output were obvious.  We installed a smart async
board (Stargate, highly recommended) and eliminated this problem.

In any event, I'd highly recommend adding status lights to any internal
modem.  This is easy to do (low current LED and a resistor) and will 
help you solve many problems that are very difficult to instrument 


John De Armond, WD4OQC  | We can no more blame our loss of freedom on congress-
Radiation Systems, Inc. | men than we can prostitution on pimps.  Both simply
Atlanta, Ga             | provide broker services for their customers.
emory!rsiatl!jgd        |  - Dr. W Williams |                **I am the NRA**  

More information about the Comp.unix.i386 mailing list