[Asterisk-Users] Digium vs. Sangoma: Performance

mattf mattf at vicimarketing.com
Thu Jun 9 10:49:30 MST 2005


Hello,

Several people asked to get a hold of the stats I used to determine that the
Sangoma T1/E1 boards performed better in our real-world tests than the
Digium boards. We've decided to post our results after confirming them over
the past month of operations. 

Here is a link to the last two reviews I wrote about Sangoma with Asterisk:
http://astguiclient.sourceforge.net/Sangoma_experience.txt

Here is a summary of the production environment that these servers were
evaluated in and the hardware used.

HARDWARE:
Each server was constructed in house with the following components:
- 4U rackmount server chassis
- Asus P5AD2 motherboard
- Intel P4 3.2GHz Prescott processor socket 775 with 1 MB L2 cache
- 2GB Kingston DDR2-533 RAM
- 3 x 160GB Maxtor DiamondMax Plus 9 SATA 7200RPM hard drives
- Enermax 550W power supply

SOFTWARE:
- Slackware 10.1 with custom kernel 2.4.29
- Asterisk 1.0.6
   the configurations in zaptel.conf, zapata.conf and extensions.conf were
the same
- astGUIclient suite
Note: Sangoma machine uses wanpipe-beta8a-2.3.3 drivers

The cards compared are the Digium TE405P in T1 mode and the Sangoma A104.

Each server is setup the same with 3 telco T1s connected(one of them
connected on the other end to another Asterisk server) and a channelbank
connected to the fourth T1 port.

The live application used to evaluate performance and gather the stats is
the astGUIclient package with VICIDIAL for outbound dialing.
http://astguiclient.sourceforge.net/

Here are links to screenshots of our performance monitoring web pages for
each server:
http://astguiclient.sourceforge.net/VDreports/performance_Digium_TE405P.gif
http://astguiclient.sourceforge.net/VDreports/performance_Sangoma_A104.gif


The data was gathered every 5 seconds from both servers.

You can see that even though the Sangoma server ran at slightly higher call
capacity(11% more calls with 4% more talk time), that it still had a
significantly lower average load(27% lower) and peak load(56% lower). The
periodic tests that we have done over the last month confirmed these
results, that all things being equal, the Sangoma server performed 30-50%
more efficiently at doing the same job.

I am currently awaiting a beta test version of the Digium TE406P card(with
the echo-canceller daughterboard) and I am awaiting word from Sangoma on
testing their newly announced A104d card(with on-board echo-canceller) to
see how they both fare in our real-world tests.

MATT---



More information about the asterisk-users mailing list