Johan,

The procedures for using the Alternate Gatekeeper fields are fully defined in the H.323v4 document and are applicable to H.323v2 systems.

You can get a copy of the most recent draft copy here: ftp://standard.pictel.com/avc-site/0008_Por/APC-1874.zip

Best Regards, Paul

----- Original Message ----- From: "Johan Gerhardsson (ETX)" Johan.Gerhardsson@ETX.ERICSSON.SE To: ITU-SG16@mailbag.cps.intel.com Sent: Thursday, July 27, 2000 11:46 AM Subject: alternateGk mechanism?

Hello all,

my name is Johan Gerhardsson and I was wondering if any one could help me

with a couple of questions regarding the of the alternateGk machanism (h.323v2).

How is it possible to make an endpoint (gw) gracefully switch over

registration and new calls to another gk without loosing existing calls on the current GK? I.e. how can the Gk tell the gw that it should send its SETUP messages (and keep alive RRQ) to another Gk?

I see two possible ways that is almost working:

1. ARQ/ARJ with altGkInfo:

What if one wants to use pregranted ARQ? Is the registration (keep alive RRQ) also applied?

2. keep alive RRQ/RRJ with altGkInfo:

What happens with existing calls on the GW if the registration is rejected

by GK?

My scenario is this: Two GK's (Gk1 and Gk2, h323v2) is together working in a load sharing mode.

Registration data are duplicated. Calls are not duplicated. The two GK knows that the endpoint Gw shall have Gk1 as primary GK.

1. Gw: RRQ -> Gk1
2. Gk1: RCF (alternateGk=[Gk2's rasaddress,<gkid>,needToRegister =

FALSE,prio=1], pregranted=yes) -> Gw

3. Gw sends keep alive RRQ to Gk1.
4. Gw sends SETUP messages to Gk1 and calls are now ongoing (gkrouted)

through Gk1.

5. Gk1 goes down (calls are cleared by the Gw).
6. Gw detects that Gk1 has gone down.
7. New call: Gw sends ARQ to Gk2 to get Gk2's call signaling address which

is saved for further use.

8. Gw routes new calls through Gk2 (no ARQ is used since pregranted is

used).

9. Gw sends keep alive RRQ to Gk2.
10. Gk2: RCF (alternateGk=[Gk2's rasaddress,<gkid>,needToRegister =

FALSE,prio=0]) -> Gw

11. Gk1 comes up again.
12. Gk2 detects that Gk1 is up again and wants to tell the Gw that it

shall go back to "normal operation" (step 3-4) as this is the wanted scenario with this load shared network setup.

Questions:

1. If pregranted was not in use, the Gk2 could use the alternateGk

mechanism with ARQ by sending ARJ with altGkInfo = altGkIsPermenent = TRUE, alternateGk=[Gk1,<gkid>,needToRegister=FALSE,prio=0],[Gk2,<gkid>,needToRegis ter=FALSE,prio=1], but the ARQ is only valid for new calls, not calls AND registration (or?).

In the case above, pregranted is wanted to reduce SETUP time, so what

shall the Gk2 do?

2. In step 9. the Gw has begun to send keep alive RRQ to Gk2. How shall

Gk2 tell the Gw to send keep alive RRQ to Gk1 when it is up and running again? (By using RRJ and altGkInfo you might think, but how is the GW handling the active calls that is routed through Gk2? Is it safe to make the assumption that the Gw will not clear these calls?)

Thanks in advance for your comments.

Johan Gerhardsson Software developer Ericsson Telecom AB Data Backbone & Optical Networks, IP telephony

Phone: +46 (0)8 719 2396 email: Johan.Gerhardsson@ericsson.com

For help on this mail list, send "HELP ITU-SG16" in a message to
listserv@mailbag.intel.com

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For help on this mail list, send "HELP ITU-SG16" in a message to listserv@mailbag.intel.com