[VOIPSEC] IPv6 and the demise (or not) ofNAT(wasRe: Interactive Connectivity Establishment (ICE))

[Mark Teicher]

Volker,

Actually it depends on the type of protocol and codecs a particular firewall supports.  Not all firewalls or edge routers support all codecs within a given VoIP Protocol (i.e. SIP or H.323).  Under ideal conditions, one protocol, one particular codecs and no background traffic, some firewalls could support up to 10,000 calls on a particular protocol and codec (most likely G.729a (small bytes, small sample rate, but makes the firewall router and edge router work very hard to pass those small packets very fast), but when one adds in background traffic, a mixture of codecs with various byte sizes and sample rates, this number will not be as high as one would expect.  Depending on the architecture(single cpu, dual cpu, use of DSPs, and various other design implementations) of the particular firewall vendor's product or edge router.  

Then your statements also do not accounting for VQ (PESQ, PAMS or other VQ quality measurements), jitter, line quality, length of call, silence suppression enable/disabled, encryption enabled/disabled, etc.  

-----Original Message-----
From: Volker Tanger <vtlists at wyae.de>
Sent: Nov 16, 2005 6:03 PM
To: Voipsec at voipsa.org
Subject: Re: [VOIPSEC] IPv6 and the demise (or not) ofNAT(wasRe:	Interactive Connectivity Establishment (ICE))

Greetings!

"Hadriel Kaplan" <HKaplan at acmepacket.com> wrote:
> As for available bandwidth, most SBCs can get to at least 2-4Gbps of
> media throughput, some much more.  The largest I've seen is 32Gbps
> (aggregate), but undoubtedly capacity will grow with the market needs.
>  Frankly, the
> access edge routers rarely have the capacity to forward that much of
> data, let alone voice.

...and firewalls usually have even less. Especially when talking about
small packets which are usual for "live" connections like voice.
"Gbit/s" firewalls often are marketing bubbles - the worst I've seen
claimed 2 Gbit/s throughput - while only being equipped with 10/100
interfaces. Rrriiight. But even the "real" Gbit/s ones rarely approach
the 1 Gbit/s limit under realistic VoIP circumstances. 

So if heading towards that order of throughput one should better think
about parallelizing/clustering and decentralizing of all systems
involved.


Let's do some rough calculation: taking for example Voice, GSM codec
(13kbit/s), SIP, RTP+RTCP, thus ~60kbit/s total bidirectional traffic
for each conversation (including overheads). Thus a 1Gbit/s (eth
interface) system will probably max out at around 10.000 simultaneous
calls ethernet-wise. Pretty much suitable for the very most companies
and campuses.

For current video telefphony (Cisco Video Link) data only is 384kbit/s,
thus with overhead probably a bit above 2 Mbit/s, thus maxing out the
Gbit-link around 300 simultaneous calls. Which is a bit limiting,
especially for bigger companies. 

Switching to 10Gbit/s interfaces will raise the numbers by one magnitude
- but 3000 simultaneous video calls are not overly excessive -
especially not if we're talking about the current upper technical limit.

Bye

Volker


-- 

Volker Tanger    http://www.wyae.de/volker.tanger/
--------------------------------------------------
vtlists at wyae.de                    PGP Fingerprint
378A 7DA7 4F20 C2F3 5BCC  8340 7424 6122 BB83 B8CB

_______________________________________________
Voipsec mailing list
Voipsec at voipsa.org
http://voipsa.org/mailman/listinfo/voipsec_voipsa.org