[Asterisk-Users] Skips and Pops in Call Recordings

Matt Roth mroth at imminc.com
Tue Dec 13 09:51:12 MST 2005


Matt Florell wrote:

 >Hello,
 >
 >Need some more information here:
 >- hardware specs (including what kind of hard drives?)
The Asterisk server is a Dell PowerEdge 6850 with the following specs.  
Please note that we are NOT recording to the hard drive.  We are 
recording to a RAM disk as detailed here 
(http://voip-info.org/wiki/view/Asterisk+Dimensioning) under the heading 
"512 simultaneous SIP-to-SIP calls with Digital Recording".  
Unfortunately, the scalability tests we did at that time assumed that if 
call quality was good, so was the quality of the recording.

Processor:  Quad Intel Xeon 3.16GHz/1MB Cache
Memory: 20 GB DDR2 400MHZ Single Ranked DIMMs (4 GB System / 16 GB RAM Disk)
Hard Drive:  Two 73GB, U320, SCSI, 1IN 15K Configured in a RAID 1 (Mirrored)
Hard Drive Controller:  Embedded RAID - PERC4 Integrated (Driver: 
megaraid_mm, megaraid_mbox)
Everything else: 
http://www.dell.com/downloads/global/products/pedge/en/PE6850_specs.pdf

 >- Linux kernel version
2.6.12-1.1376_FC3smp (Fedora Core 3).

 >- running Xwindows?
No.

 >- Asterisk version
ABE-A.2-beta (Asterisk Business Edition A.2 beta).

 >- kind of calls you are recording (Zap, SIP, IAX, Meetme, ...)
Calls originate on the PSTN and are handled by a Cisco AS5400 Universal 
Gateway that is a SIP peer of Asterisk. The AS5400 converts the calls 
from TDM channels to VoIP (SIP) channels before sending them to 
Asterisk.  The Asterisk dialplan then routes them to one of our agents, 
who are using SNOM 320 VoIP (SIP) phones.  Essentially all of our calls 
are SIP-to-SIP, with absolutely no protocol bridging or transcoding 
occurring on the Asterisk server.  

The Asterisk server handles the following major tasks:

- Routing calls through the dialplan to (dynamic) agents in the 
appropriate queues.
- Adding/removing agents to/from queues via AddQueueMember and 
RemoveQueueMember (NO static agents!).
- Recording calls via the Monitor application directly to RAM disk. 
Calls are moved to a remote machine for mixing.
- ChanSpy-based quality assurance of calls.  Neither ChanSpy nor the 
quality of the calls themselves is affected by the problem.

 >- how many recordings at once
Anywhere from 5 to 30 concurrent recordings.  This is not our planned 
peak, but it's where we've experienced the problem so far.  We have not 
yet determined if the number of concurrent recordings is an issue, but 
we are considering it.  We also haven't determined if the problem gets 
worse as the number of recordings increases, but it definitely exists 
throughout that entire range.

 >In my experience, HyperThreading does not cause recording problems,
 >it's usually a disk issue. When we had issues, switching to fast SCSI
 >drives on a MegaRAID 320-1 with the megaraid2 linux driver solved all
 >of our problems(skips and clicks/pops)

The disk issues also directly interfere with call quality, as our 
previous scalability tests showed.  Digium seems to think that the issue 
is scaling (some resource contention that causes a bit of audio to be 
unavailable when the write occurs).  I see their point, but given our 
hardware and the current call volume I'm not completely sold on it.  
Could it be a configuration issue (file handles, interrupts, etc...)?

 >MATT---

Colin Anderson wrote:

 >Matt Roth: FWIW, I am recording 1000-1500 calls a day (as of 8:52PM today,
 >1482 calls!) of various length on my Netfinity with the onboard IBM RAID
 >controller in RAID 5 Ultra 320 SCSI with suprisingly good quality. As the
 >other Matt indicated, maybe what is needed here is an intelligent 
controller
 >to offload some of the chore.
 >
 >No definite solution here, but at least it's another data point to 
compare.

I appreciate any information contributed by list users. It's by far the 
most valuable resource available to me.

 >On 12/12/05, Matt Roth <mroth at imminc.com> wrote:
 >
 >>List users,
 >>
 >>I'm using the Monitor application to record calls.  Most of the
 >>recordings are audible, but contain skips accompanied by a popping
 >>sound.  Sometimes they are isolated, sometimes they appear in groups.
 >>Call quality is excellent and seems unaffected by whatever is causing
 >>this problem.
 >>
 >>If anyone has experienced this problem before, I'd appreciate if you'd
 >>share what the source was and any tips on eliminating it.  I contacted
 >>Digium tech support and they suggested turning off hyperthreading.  I
 >>have done that, but I won't know if it improved things until tomorrow.
 >>
 >>The machine is running at a moderate call volume and is always at least
 >>90% idle.  I'm not seeing any "Avoided deadlock" messages in the logs.
 >>If you need any more information, I'd be happy to provide it.

Thank you,

Matthew Roth
InterMedia Marketing Solutions
Software Engineer and Systems Developer



More information about the asterisk-users mailing list