X.25 PAD as a local multiplexor for UNIX tty lines?

[David Sherman]

We have a fairly complex communications setup, part of which includes:

Users in Ottawa                                               Memotec
and other cities --- X.25 PAD --> packet-switched network --> X.25 PAD
                                                              at host in
             Administrative machine #1 <--- Gandalf PACX <--- Toronto
                                                  |
                                                  v
       Gandalf 309 short-haul modem <-- 4-channel multiplexor
         |
       leased line to second office 2 blocks away
         |
       Gandalf 309       ---->   4-channel     --->  Administrative
       short-haul modem          multiplexor         Machine #2


In other words, the PACX allows users in our remote offices to
sign on to either administrative machine, one of which is in the
machine room with the PACX and the other a couple of blocks away.
It all works fine (and in fact I'm simplifying a lot of detail).

Due to increases in out-of-town staff, we need to increase both the
number of ports on our Memotec PAD (from 16) and the number of PACX
channels.  As an alternative to simply upgrading, we're looking seriously
at a new box from Gandalf called an "Access Server", which combines
the functionality of PAD and PACX.  (Anyone have experience with this?)

Our original plan was to also replace the 4-channel multiplexors
(Gandalf PIN 9106, quite old) with two 8-channel ones (Gandalf Mini-Mux),
since more ports are needed into Admin Machine #2.

The Access Server becomes cost-effective if we can use the Memotec
PAD, which will now be surplus, as a multiplexor rather than buying
two Mini-Muxes.  The Access Server, with its X.25 capability, would
play the role of multiplexor in place of the 4-channel mux at the
machine room end.

Now, my question: I've used PADs and X.25 quite a lot, but always
over a packet-switched network, where packet turnaround time is
slow, and packets are typically sent when full, on a timeout or
(from an interactive user) on a CR or other control character.

If we want this PAD to work as a multiplexor, however, we'll have
to configure it to forward a packet on every character, with no
delays.  Otherwise UNIX tty echo won't work, and of course screen-based
editors would be useless.

On a local leased line supported by short-haul high-speed modems
(19200 bps), can a PAD packet and depacket X.25 fast enough to
make it work transparently as a multiplexor?

Any comments appreciated.

David Sherman
Computer Education Facility
The Law Society of Upper Canada
20 Queen St. West, 34th floor
Toronto, Canada  M5H 2N6
(416) 947-3466 (at LSUC); (416) 889-7658 (most days)
-- 
Moderator, mail.yiddish
{ uunet!attcan  att  utzoo }!lsuc!dave          dave at lsuc.on.ca