11 Feb
2005
11 Feb
'05
1:54 a.m.
I have to admit I'm not so much involved in the theoretical discussions.
But from practical point of view, I see no difference on doing the NAT
traversal on the main server or remotely (via SBC) - you still have to
do some media relaying, so QoS and delay will be affected in the same way.
For a worldwide spread VoIP provider, having a centralized NAT traversal
will not scale, risking a bandwidth exhaustion because of media
relaying. We came up with a solution based on a distributed set of NAT
traversal server, configuration which will eliminate bandwidth
bottle-necks and the existence of the single point of failure. Even
better, you will get a better QoS since the media doesn't have to cross
half of globe just for the sake of NAT traversal.
We can also see this approach in Skype design and this is one of the
think that contributes to its success.
Now, Jiri, don't get me wrong, but I didn't suggest to use SBCs in order
to achieve High-Availability. It was just sharing with Darren some of
our experience in deploying High-Availability support for SER platforms
vis-a-vis of any type of NAT traversal - local or remote.
Best regards
Marian
Jiri Kuthan wrote:
There is a recent discussion about SBCs on the sip
forum mailing list.
shortly, sbcs are a technique to harm QoS through bandwidth consumption
and packet latency, and to affect reliability through introduction
of a single point of failure. There are also extensibility concerns.
One can still achieve NAT traversal and failover capability without
SBCs. Which is not a blank statement -- my company, iptelorg, has developed
such product recently.
-jiri
At 11:41 PM 2/10/2005, Marian Dumitru wrote:
>Hi Darren,
>
>Disregarding any implementation aspects, the only complication a SBC can introduce is
an additional hop in the signaling path.
>On the other hand, the SBC comes into focus when is about:
> -decoupling the NAT traversal from the routing logic - in case of a very
complex service and routing logic or when is about considerations like yours;
> - distributed NAT traversal - keeping the media as local as possible in
platforms with a wide-geographical coverage.
>
>Best regards,
>Marian
>
>Darren Sessions wrote:
>
>>I sent the email to the mailing list and realized the answer about 15
>>minutes afterwards. Your email Jan, confirms it.
>>I had discussed session border controllers with Jiri many months ago and was
>>told a session border controller was not a good approach as they severely
>>complicate signaling matters.
>>Other than using a session border controller, are there any viable solutions
>>to this problem without resorting to a IP failover cluster or something of
>>that nature?
>>Thanks,
>>- Darren
>>
>>On 2/8/05 5:49 PM, "Jan Janak" <jan(a)iptel.org> wrote:
>>
>>
>>>No, because RTP proxy would relay media only. SIP signalling would still
>>>go through one of the proxy servers and SIP messages would only make it
>>>to the user agent behind symmetric NAT if they were sent by the proxy
>>>server originally contacted by the user agent (with the same IP address).
>>>
>>>Jan.
>>>
>>>On 08-02 13:19, Darren Sessions wrote:
>>>
>>>
>>>>We currently do not use an RTP proxy in our service (so the audio does
not
>>>>ride our internet bandwidth).
>>>>
>>>>Our biggest issue at the moment is the redundancy between two SER servers
in
>>>>dealing with symmetric NATs (specifically dealing with the individual SER
>>>>server unique IP addresses and the far end customer's symmetric NAT).
>>>>
>>>>If we were to use an RTP proxy, as a backup mechanism for dealing with
NATs,
>>>>would this alleviate the issue of multiple SER servers and symmetric
NATs?
>>>>
>>>>
--
Voice Sistem
http://www.voice-sistem.ro