flexfax: Hylafax success story / DNIS routing

[Alan Wagner-Krankel <awkwardly@yahoo.com>]

With all of the recent list traffic on DNIS/DID,
this seemed like an appropriate time to describe
our recent (and mostly successful) Hylafax
implementation.

Our system is used only for incoming faxes - we
have a pool of 48 modems, and 17 numbers assigned
to the pool.  An incoming call on one of the 
numbers is assigned to any of the 48 modems -
faxes are routed to the appropriate destination
based on the DNIS digits (the 'how' is described
below).

As you can probably guess from the use of the
acronym 'DNIS', our equipment sits at the end of
a pair of T-1s, and acts like a PABX.  The other
end of the T-1s is our switch, a Nortel
Meridian SL-100.

The modem equipment is a 3Com/USR Total Control
rack.  It has a T-1 card that acts as the CSU/DSU,
a dozen 'Quad' modem cards, and another card which
provides SNMP services.

Calls come into the T-1 card, and are sent along
a midplane bus in the rack to one of the modems.
The modems see it just like an analog call.

When extended result codes are set on the modems
(ATX7), the 'RING' result code includes the
DNIS digits, e.g. 'RING / 54321'.

By default, Hylafax does not handle this extended
RING result code - no surprise here, since AFAIK
this is purely a Total Control invention.  I
modified the code to parse out the DNIS digits, 
and then look up routing information in a
simple ascii table.

At the moment, my code changes are pretty ugly -
this was one of those times the prototype became
the release.  As such, I haven't try to package
them as something to be incorporated in the main
release - besides, this is probably not a very
general problem or solution.  If there is interest,
I might to be able to create a more detailed
how-to document.

Some details (somebody may care):
Hylafax: 4.0pl2, with modifications
Server: AIX 4.3
Modem rack: 3Com Total Control
    modems: Quad V.34, firmware 6.0.6/6.1.6
    T1: Dual Channelised T1 (186), firmware 3.5.0
    SNMP: Network Management Card, firmware 6.0.9
    All these are part of TCS version 3.5.  Note
    that I had some T1 problems that went away
    with these firmware revs.
T1 parameters: ESF/B8ZS, DNIS tone type is DTMF


Anyway, the system is working well - thanks to
the maintainers, and to the folks who have been
keeping the mailing list going.  This software
has proven invaluable.


I have one problem that I may as well pose to
the list while I'm here:  I am getting a fair
number of 'Ring detected without successful
handshake' errors - some days 10 percent of
the calls fail this way.  Does anyone have a
feel for what this percentage should be? Are
my modems being too picky about line conditions
and such, or are they just bad? (they are USR's,
after all.)

Thanks,
Alan Wagner-Krankel

_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com