[asterisk-users] Ast/Hyla/IAX Scalability?

[Brent Davidson]

David Backeberg wrote:
> On Sat, Mar 14, 2009 at 12:00 AM, Steve Underwood <steveu at coppice.org> wrote:
>   
>> Fully open-to-the-public FAX servers tend to get just get a lot of bad
>> calls, many of them wrong numbers, or voice users. FAX servers for
>>     
>
> I've definitely seen that, and have been able to either identify the
> validity of a caller by CID or by calling the number and confirming a
> blast of fax tones.
>
>   
>> clue what kind of failure rate might be expected. You can find a bit
>> more about these issues and our results at
>> http://www.soft-switch.org/spandsp-soft-fax-performance.html
>>     
>
> After reading that, it occurred to me that I'm running SpanDSP 0.0.5
> and 0.0.6 seems to have enhancements that may solve the problems I've
> been seeing. I'm convinced that it's worth upgrading and seeing if I
> can reduce my failure rate.
>
>   
>> Your differing failure rates between using ReceiveFAX and using iaxmodem
>> seem to indicate your results relate to issues in your own system,
>>     
>
> I think I wasn't very good at setting it up, as I had no experience
> with IAX. Likely my fault rather than anything inherently wrong with
> the software. There were more moving parts than I was able to get a
> handle on, and when I switched to 1.6 and app_fax things 'just
> worked'. This is why I keep recommending the 1.6 approach over the 1.4
> + IAX + IAXModem + Hylafax.
>
>   
>> LANs don't loose packets), will have a true failure rate (i.e. a rate of
>> calls failing which had the potential to succeed) well below 1%. The
>>     
>
> That's consistent with my testing before I set it live.
>
> You mentioned recording faxes. I know how to do that with IAXModem,
> but are you familiar with a method for 1.6 and app_fax? I read through
> app_fax.c and didn't see any way to send a flag. Is the recording
> built into SpanDSP, or is is something IAXModem added on themselves?
>
>   
For what it's worth, the company I work for switched from WinFax to 
HylaFax last spring.  We only have 4 analog phone lines coming in to a 
4-port modem card, but the Hylafax system runs on the same server as our 
main Asterisk PBX.  So far Hylafax is performing much better than WinFax 
ever did.  When we have errors either sending or receiving, it is always 
either line problems or the wrong number being dialed resulting in a 
voice call to the fax line.

I would estimate that our overall success rate is around 95% if you 
disregard faxes to wrong numbers or incoming voice calls to the fax 
lines.  Load testing a large-scale fax system under real-world 
conditions is difficult if not impossible without having access to a 
variety of hardware and software fax devices scattered all over your 
prospective send or receive area.  If you load test from your own 
location by attaching a bunch of fax machines or a fax sending server to 
your outgoing lines and have them dial back in, then you're only looping 
through your local telco's switching center.  You might get very 
different results from sending faxes from out of state, or even across 
town.  It's been my experience that telephone line quality varies 
greatly from place to place and even from time to time.

A perfect example is from back in my days as a systems admin for a 
dial-up ISP.  We were operating in a small town where PRI or channelized 
T1's weren't available so we had a bank of about 100 US Robotics 
external modems connected with serial cables to 2 Livingston PortMaster 
terminal servers.  Everything would run fine (or as fine as it ever got 
with dial-up) until it decided to rain.  Everytime we'd get more than a 
tenth of an inch of rain a large group of the modems would go haywire 
and start dropping calls.  A couple of the modems would burn out 
completely.  We had the telco out repeatedly and they always gave us 
some answer that didn't make any sense.  After about the 6th time this 
happened they sent out a technician with a brand new line analyzer that 
happened to include a TDR.  The vast majority of the lines we were 
having trouble with showed to have a partial short about 100 feet from 
our building which just happened to be right under the middle of the 
road in front of our building.  They dug the section of line up and 
found that the cable had been partially cut at some point in the past 
and the wires were spliced with electrical tape and the whole bundle had 
then been wrapped with tape.  Every time it rained, the water would seep 
into the shoddy splice and short all the lines together.  When the water 
dried out, the shorts would go away and the lines would go back to normal.

I've seen situation like that enough to know that until everybody has a 
purely digital phone line, there will always be line quality problems 
that will be out of the end user's control.  Even though the company I 
work for now is a small company is a very rural area where technology is 
somewhat limited, we're beginning to realize just how antiquated Fax is 
becoming.  E-mail and web services are rapidly replacing fax to the 
point that 90% of our incoming faxes are spam.

One really nice thing about Hylafax is the logging that happens during 
the entire fax process, so it's much easier to see where the problem is 
occurring.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.digium.com/pipermail/asterisk-users/attachments/20090316/4f0dea58/attachment.htm