Skip to main content

VMWare Audio Quality

Posted by Phoneguy on Wed, 04/19/2017

We are testing MTE on vmware with 4 x CPU. We are getting audio glitches from time to time. We found that the biggest problem was related to munin running something every 5 minutes so we stopped that process and things improved. It still seems to be happening in particular when the perl scripts are running when the website is reloading. Use of 20-30% of just one of the cpus causes a brief audio blip. Logging in to the website is enough to do this. The total cpu usage never goes over like 10%

Anyone else seen similar problems with vms, and ideas on a solution?


Submitted by volodya on Wed, 04/19/2017 Permalink

Hello,

Do you have other VMs running on the same hypervisor? Can you check overal system performance and cpu iowait values for your PBX VM in particular?

Submitted by Phoneguy on Wed, 04/19/2017 Permalink

It is a shared cloud so we don't have access to what else is going on on the same hypervisor. Maybe that is the problem in itself. In the vsphere web client, we are seeing disk latency of 5-15ms which our provider claims is normal.

Submitted by netriplex on Thu, 04/20/2017 Permalink

We have multiple instances of Thirdlane MTE and STE running on our Openstack based cloud infrastructure and do not have this issue.

As alluded to by volody, this is most likely an issue with the performance of your providers cloud infrastructure.

Submitted by eeman on Thu, 04/20/2017 Permalink

the reason VMs sometimes have issues is due to the lack of a true timing source. VoIP is heavily reliant on timing especially when encoding the audio from untimed sources like playback of audio from a file, conferencing, and when mixmonitor is invoked. A busy server with multiple guest domains has to timeshare cpu clock cycles with other guest domains and can result in cpu lag that can have this result.

Submitted by netriplex on Fri, 04/21/2017 Permalink

eeman, as you said, a "busy server" has these issues, however, if the host I snot overloaded this has never busy a problem for me.

We have over a dozen instances of Asterisk/Thirdlane (some Thirdlane, some Vanilla Asterisk) and have not had these issues.

phoneguy,

We run our own openstack but we are a commercial cloud provider. However, we never load our nodes above 50% on average which is probably why we do not have issues. This also makes our pricing a little higher than the $5-$10 per month cloud providers though.