Re: RFC 4666 M3UA - Reg Request Message

Ilie,

Discarding messages is the wrong thing to do.  It appears that you have not had
much recent experience with designing and fielding ISUP systems.

ISUP systems that distribute call handling capacity by CIC ranges must still
share circuit group supervision state across the architecture.  Otherwise,
circuit management would not be possible if a call processors for a CIC range
failed.  Therefore, regardless of the failure of a call processor, the system
must still maintain circuit supervision state.  Poorly defined systems that
fail to do so result in an annoying circuit leakage.  Therefore, at least an
administrative processor must be capable of obtaining and maintaining circuit
management state for every circuit.

An ASP in the system is in the best position to obtain this circuit management
state for any given circuit.

Therefore, if an SG (albeit one within the MGC complex) pends on a CIC range
and is forced to moved to the LG-INACTIVE state for the load group,
LOADSEL/LOADGRP allows it to select an ASP (or ASPs) from the available ASPs
serving the AS an distribute messages related according to the traffic mode
defined for the AS.  If that traffic mode is load sharing, it will distribute
these unserved CICs across the ASPs that are available for the AS, regardless
of which Load Selection for which they are active.

An ASP serving the AS is thus able to institute appropriate software failure
blocking procedures for circuits for assigned circuits for which it receives
messages and is also able to properly complete blocking procedures because the
SG will also deliver subsequent messages related to the same CIC to the same
ASP.  Each ASP needs to be capable of accessing the centralized circuit
management state for each circuit, but for proper circuit management as
outlined above, this is necessary for a properly functioning system anyway.

Thus, by delivering unserved load selection messages to the ASPs, the remote
switch responsible for the same circuit range will block call to the circuits
and take appropriate action with regard to alarms, network management controls
(call rerouting), etc.  This is far preferable to throwing messages on the
floor and waiting for call processing related timers to expire, just to have
the remote switch reselect the circuit when they do.

--brian

Ilie Glib wrote:                        (Thu, 09 Nov 2006 16:57:10)
> Lincoln,
> 
> if AS PENDING state changes to INACTIVE, and there are still ASes
> serving the same SPC and SI, the SG can do nothing but discard
> buffered messages.
> This is currently suggested by the SIGTRAN RFCs, I think it is
> reasonable (there is also the default AS concept, but IMHO it does not
> work for CICs since there is a need to coordinate CIC states at the
> ASPs serving the AS).
> 
> Regards
> 
> /Ilie
> 

-- 
Brian F. G. Bidulock
[email protected]
http://www.openss7.org/