[SR-Users] Kamailio 5.0 - B2BUA

[Alex Balashov]

Daniel,

(Yes, it was addressed to you. :-)

On 03/23/2016 02:43 PM, Daniel-Constantin Mierla wrote:

> I haven't rejected any patch with new features in kamailio that
> doesn't have significant impact on performance and doesn't break
> existing and expected functionality. So anyone is welcome to bring in
> code ...

Fair enough. It's just that the number of people with sufficient SER
core development expertise to do this is, realistically, I think, quite
small. I certainly couldn't do it.

> Otherwise, besides (hopping to) getting the new topos module in good
> shape as a topology (stripping) firewall, I have no business plans
> that would push me to invest resources in a b2bua.

Understood. What is the essence of the new 'topos' module? I am not sure
I have understood it.

> You said sems is calable but lacks docs, I guess that is less time
> consuming to sort out for someone really interested given it is open
>  source than implementing something from scratch.

I don't think it lacks docs so much--the Readmes embedded in the repo
are actually pretty good. And it definitely performs well under high
loads in my observations, especially with the SIP thread pool enabled,
as discussed in Stefan's KW presentation some years ago.

What it lacks unfortunately is what might be termed FOSS critical mass.
The maintaining company swamped with work on their commercial endeavours
and development of their flagship commercial product (understandable).
The SEMS list is dead, apart from very occasional shot-in-the-dark
questions answered when Stefan has limited time. A lot of the answers
are kind of "go hack this in the source yourself", which is
understandable too -- it's not a criticism at all. Frafos have made
clear (from my vantage point) that they are not even especially
interested in supporting open-source SEMS on a commercial basis per se.
Perhaps I am wrong about that, though, and if anyone from Frafos would
like to weigh in on that and correct me, please do.

So, mostly empty mailing list, no thriving user community, conferences,
or an ecosystem of O'Reilly and Packt Publishing books. You can
understand why someone trying to use it seriously as critical
infrastructure might look at this as a nonviable avenue for commercial
support backstop and long-term technology strategy.

I myself am an advocate of strongly pushing SEMS as the official
companion product to Kamailio for B2BUA applications rather than
building a B2BUA in Kamailio. I also think that could serve as a useful
commercial pipeline to Frafos for the ABC SBC product, some percentage
of such users would surely want to 'upgrade' to the 'real deal'. But I
myself don't of course have the resources or time to become The SEMS Guy
for these purposes, and neither do most individuals.

There's a kind of spontaneous criticality effect in open-source that
materialises around certain projects and not others, and so far that has
not materialised around SEMS, although perhaps I am ignorant of
widespread, quiet utilisation.

The only major uses of it in larger commercial platforms of which I am
aware are that of Sipwise, Juha Heinanen/Tutpro's OpenSIPg product, and
our CSRP. My guess is that both Sipwise and Tutpro have separate
commercial arrangements with Frafos for SEMS issues, so most stuff
around open-source SEMS stays in the shadows.

If that can change, I think their engine is definitely the most natural
and complementary fit for Kamailio for these use-cases. But it's not 
clear who is in a position to drive such change unless it became a 
central ingredient of Frafos' strategy per se.

> Also Asterisk or FreeSwitch can scale pretty well if time is invested
> to optimize their config, besides using dispatcher to build a farm of
> them...

The general problem with this is economic in nature rather than 
technical. Freeswitch is prone to deadlock around 300 CPS, so if you 
want to handle, say, 2000 CPS, as is not uncommon in the short-duration 
industry and also in larger "legitimate traffic" Installations, you need 
a farm of 5-10 Freeswitch boxes. That's a lot of boxes, and boxes are 
expensive. Kamailio can otherwise handle such throughput 
unproblematically on one server. I don't know how Asterisk call setup 
performance is these days, but I suspect it is not better than Freeswitch.

-- Alex

-- 
Alex Balashov | Principal | Evariste Systems LLC
1447 Peachtree Street NE, Suite 700
Atlanta, GA 30309
United States

Tel: +1-800-250-5920 (toll-free) / +1-678-954-0671 (direct)
Web: http://www.evaristesys.com/, http://www.csrpswitch.com/