vicidial.org

What can cause degraded sound quality that makes it difficult for customers to communicate effectively with agents? The server works fine after a new installation, whether it’s a scratch install or using the ViciBox OS. However, after a few hours or a day, the sound quality deteriorates. When we reinstall, the sound quality returns to excellent. Do you have any suggestions on what might cause this issue?

is this a single server or a cluster? Are you using an old or backed up database when you reinstall? if so, how large is the database? whats the load on the server(s)? how many agents? what dial ratio?

carpenox wrote:is this a single server or a cluster? Are you using an old or backed up database when you reinstall? if so, how large is the database? whats the load on the server(s)? how many agents? what dial ratio?

Yes, it is a single server. No backed up database.

Currently,

-> FROM information_schema.tables
-> GROUP BY table_schema;
+--------------------+--------------------+
| Database Name | Database Size (MB) |
+--------------------+--------------------+
| asterisk | 48.57179260 |
| information_schema | 0.20312500 |
| mysql | 3.35156250 |
| performance_schema | 0.00000000 |
| sys | 0.03125000 |
+--------------------+--------------------+
5 rows in set, 26 warnings (0.144 sec)

Load is 1%, Same isssue with single agent whether dialling manually or auto dial.

Server Configuration:
CPU - AMD EPYC 4244P - 6c/12t - 3.8 GHz/5.1 GHz
RAM - 32 GB 5200 MHz
Data disks 2×960 GB SSD NVMe (Soft RAID)

Try another carrier next time instead of reinstalling. Note that if the problem is bandwidth at your location, reinstalling could often take enough time to allow usage of others you can't see or know about to decrease... making you think the reinstall solved a problem. If it's the carrier's bandwidth or server load, changing to a new carrier could resolve the issue (always have at least two, preferably three!).

If the problem is bandwidth at your location (your ISP or your building sharing bandwidth ...), you can try getting second connection or switching to a low-bandwidth CODEC such as GSM or g729. G729 could heavily load your CPU and require multiple dialers for the compression routines, but does dramatically reduce bandwidth requirements.

Is your CPU usage during these "bad quality calls" heavy? What average server load (from the command "uptime" at the CLI) do you have DURING these issues (not a few minutes later ... DURING)? Also consider turning off hyperthreading (could make it better or worse, we've had it go both ways).

I'm not a fan of software RAID, but since you're in nVME territory you have no choice. If all other roads are dead ends, however, you might look at removing the software raid. I doubt that's it, but ...

Hello William,

There are no issues with the bandwidth. The CPU load is normal. Tried with Hyperthreading on/off. After searching online, we found kernel-rt package which is good for real time applications. Not sure if that can resolve the problem.

williamconley wrote:Try another carrier next time instead of reinstalling. Note that if the problem is bandwidth at your location, reinstalling could often take enough time to allow usage of others you can't see or know about to decrease... making you think the reinstall solved a problem. If it's the carrier's bandwidth or server load, changing to a new carrier could resolve the issue (always have at least two, preferably three!).

If the problem is bandwidth at your location (your ISP or your building sharing bandwidth ...), you can try getting second connection or switching to a low-bandwidth CODEC such as GSM or g729. G729 could heavily load your CPU and require multiple dialers for the compression routines, but does dramatically reduce bandwidth requirements.

Is your CPU usage during these "bad quality calls" heavy? What average server load (from the command "uptime" at the CLI) do you have DURING these issues (not a few minutes later ... DURING)? Also consider turning off hyperthreading (could make it better or worse, we've had it go both ways).

I'm not a fan of software RAID, but since you're in nVME territory you have no choice. If all other roads are dead ends, however, you might look at removing the software raid. I doubt that's it, but ...

more than likely your carrier then

On the recordings in your system: Are the leads and/or agent voices recording properly? Obviously if the recording of the agent is clean and clear and the Lead is not hearing "clear" we look in a completely different direction than we would if the recordings are also poor quality.

Describe the "poor quality"? Garbled? Robot sounding? Breaking up? Static? Low volume?

When you say "few hours or days" does this problem always occur with a specific minimum of agents on the server(s) where quality is low? (Note: If this is a cluster, I refer to sound quality vs number of live and ringing calls on one specific server at a time).

What is the Server Load (available easiest with "uptime" on the command line) of the offending server when this is happening? (Not 5 minutes later, not 1 minute later, but DURING)