[Users] Natping in "dispatcher scenario"

Helge Waastad Helge.Waastad at smartnet.no
Wed Feb 8 21:20:35 CET 2006


Hi,

It sounds reasonable to "dispatch" the nat ping function from the core due to load.
I'm trying to load share my clients based on DNS SRV and spread them over several OBP's.

In my scenario I'm only ping'ing UA's without keep-alives implemented (flag'ing based on UA info in header). Fortunately these are not so many. That is why I wanted to see if the scenario I mentioend was possible.

Looking forward to the updates.

Thanks for your reply.

Br,
Helge Waastad
Senior Engineer
Smartnet
tlf: 67830017



-----Original Message-----
From: Andreas Granig [mailto:andreas.granig at inode.info]
Sent: Wed 2/8/2006 8:55 PM
To: Helge Waastad
Cc: users at openser.org
Subject: Re: [Users] Natping in "dispatcher scenario"
 
Helge Waastad wrote:
> Does anyone else find it interesting to have the ping also relayed via
> the OBP?
> If not, I just have to move the registrar more towards the frontend.

We ended up in implementing a simple stand-alone natping application for
a proof-of-concept design using raw sockets for spoofing the OBP
address, mainly for two reasons:

- Performance: a possibility would be to evaluate the Path uri in the
registrar and send a sip-based nat-ping (OPTION) with a pre-set route,
but this would really stress the OBP (assume you've to ping 50.000 UACs
every 60 seconds).

- Simplicity: it would require quite a lot of changes in
registrar/usrloc/nathelper/mediaproxy, so we just query the
received-column of usrloc from the external natping application(s) and
send empty udp packets.

Patches to the registrar and a new path module for accomplishing this
(inserting proper Path headers in an OBP, and taking the
received-address from a Path uri parameter and saving it into usrloc in
a registrar) will be contributed the next days...

Bye,
Andy

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kamailio.org/pipermail/users/attachments/20060208/54491d58/attachment.htm 


More information about the Users mailing list