Folks, I'm trying to puzzle out what I can do in an IPCC design to "get around" the one-GateKeeper-per-Call-Manager-cluster rule.
Specifically, if the standard design model uses the PSTN as a backup if a GK rejects a connect message from the CM, and the standard backup is the PSTN, it seems logical that I could use a DGK to reroute to another GK elsewhere in the network if I have alternate paths to the same site.
Further, from a semantic point of view, it's inconceivable to me that the GK (eithe directly or via the GKTMP SDK and API) will only provide an RAI message that only states "bandwidth almost gone"...instead of the actual bandwith currently running through the GK. I should be able to include this bandwidth usage in my ICM routing decisions, so that I can avoid a site that is out of "trunk" space.
I should be able to redirect within the WAN to another GK (or use that GK as a conduit to a DGK) and I should also be able to determine how much of the GK's bandwidth is currently in use...whether through the GKTMP NIC to ICM, or some other "back door" method (SNMP messages/MIB traps?)
Any thoughts from the group? I'm having a terrible time (from an engineering standpoint) explaining to clients that the GK is a single point of failure, and that the only alternative backup route is via PSTN. The client basically has to provision connectivity twice between their sites--data and voice.
Thanks all for your suggestions.
