Choppy IVR and Recordings

[tsukraw]

I am hoping someone will have some ideas as to what the issue is.
We are experiencing a issue where call quality with any recording is horrible. If we call into a IVR or checking voicemail or making a conference the recording is very choppy.
The server it is running on has more then enough resources and we do not see a CPU spike when these calls are made. The system has more then enough RAM free as well. It is running on server 2008R2 which is 64bit of course.

We do have QoS setup on our server switches and phones.

I have ran a wireshark capture on the server its self and we can hear the audio issue there as well which would be before it is even on the network which should rule out Phones and QoS possible issues.

I have attached the wireshark with a call into into the conference extension 700 where you can hear the audio issue. Change the txt to a pcap and play it in wireshark.

Any ideas?

This is a single core processor i know there were issues in the past with multi core. Not the case here.
It is running onto of VMware ESXI server.

 

[lneblett]

Not sure, but I do seem to recall where some had reported a similar issue when using SSD and/or SAN for drives. The drives did not have the transfer rate to service all requests seamlessly.

 

[leejor]

I also seem to recall that (someone discovered) there was a setting in the (their) BIOS (for the harddrive) that when changed, greatly improved the recorded audio playback/recording).

 

[tsukraw]

Like i said this is indeed running on a VMware server. The storage is actually on a HP SAN.
I am giong to try to move the VM over to local independent storage to see if that improves the performance.

Is there any recommendations as to how the VMware is setup? I am fairly new to running phone systems in a virtual environment so any pointers would be great!

 

[lneblett]

I have one system running using VMware with Sans on a Synology dual gigabit device running raid5. I have no issues, but I also am running multicore with 2cores dedicated to 3cx. The system is running on Win7 Ultimate.now that Leejor mentioned it, I too seem to recall a similar item being mentioned. It wasn't that long ago and it may be worth your while to o a search to see if something pops up. I will also take a quick look.

 

[lneblett]

Take a look at attached link......maybe?

http://www.3cx.com/forums/poor-voicemail-quality-through-phone-24268.html

 

[tsukraw]

That post sounds exactly like the issue we are seeing. We are positive we have enough system resources to run this. The host OS has a single Core as recommended with 8GB ram. We only have 20 phones so it shouldn't be that intense of a setup. We are running a multi-pathing SAN that has Qty:8 15K SAS drives so we know we have enough IOPS and within VMware performance monitor we are no where near pushing the performance abilities of our physical hosts. The physical host has 24 total cores. 96GB ram and only running total 15VMs. Not sure if it at all would matter but they are AMD servers.

Has anyone every setup Scheduling Affinity? It is under Resources / Advanced CPU.

My thoughts are that i have 24 total cores and those 24 cores are being used between the 15VMs. Since the VM host is not directly assigned to a single core the timing is somehow getting off as it moves between the cores.

I might be way off on this but it is just a thought? Anyone have any input on this idea?