[jitsi-users] tunes JVB parameters in VM infrastructure


#1

Hi Jitsi Meet Community,

We're running our jItsi-meet instance in Virtual environment (VMware) and I have some question regarding OS tuning for our VideoBridge servers.
People from Nubomedia project write a good article about OS tweak for a WebRCT media server running on Virtual Machine : https://webrtchacks.com/webrtc-media-servers-in-the-cloud/.

Since April VideoBridge package installation change default network read buffer (https://github.com/jitsi/jitsi-videobridge/blob/master/config/20-jvb-udp-buffers.conf) and our server UDP loss rate have decrease since.

We're still facing some packet loss issues degrading media quality and I want to determine if it could be caused by our server or not.
Is any one running his Jitsi-Meet instance on Virtual infrastructure have try to tune other OS network parameters like in Nubomedia article ?

Best Regards,
Damien.


#2

Hi Damien,

Hi Jitsi Meet Community,

We're running our jItsi-meet instance in Virtual environment (VMware) and I have some question regarding OS tuning for our VideoBridge servers.
People from Nubomedia project write a good article about OS tweak for a WebRCT media server running on Virtual Machine : https://webrtchacks.com/webrtc-media-servers-in-the-cloud/.

Since April VideoBridge package installation change default network read buffer (https://github.com/jitsi/jitsi-videobridge/blob/master/config/20-jvb-udp-buffers.conf) and our server UDP loss rate have decrease since.

We're still facing some packet loss issues degrading media quality and I want to determine if it could be caused by our server or not.
Is any one running his Jitsi-Meet instance on Virtual infrastructure have try to tune other OS network parameters like in Nubomedia article ?

Yes, we run meet.jit.si on AWS.

The first thing to determine is where the packets are lost: in the network, or in the local kernel's buffers? Look at "packet receive errors" in the output of "netstat -asun".

If the packets are dropped in the kernel, you can check the jitsi-videobridge logs to make sure that the changes to the buffer size are taking effect. You should see something like this:
Initialized AbstractUdpListener with address 192.168.0.1:10000/udp. Receive buffer size 10485760 (asked for 10485760)

If they are, you may consider increasing them further, or using a different type of virtual machine.

Regards,
Boris

ยทยทยท

On 01/12/2017 04:47, Damien FETIS wrote: