Openmeetings Optimization

Question

open source

Please share your experience of tuning Openmeetings.

There is such a server with CentOS 6.4. on board. It has Openmeetings 2.1.1 installed with MySQL backend.

OM is started as is via red5.sh, only the following parameters for JVM are added: -server -Xmx4096M -Djava.net.preferIPv4Stack=true -Xverify:none

JVM itself:

java version "1.7.0_45"
Java(TM) SE Runtime Environment ( build 1.7.0_45-b18)
Java HotSpot(TM) 64-Bit Server VM (build 24.45-b08, mixed mode)

Two scenarios are of interest:

1. There are 15-20 people in the room, some (preferably many) have video-audio turned on. The room is loaded as an iframe on the moodle page through the appropriate plugin for the latter - but I suppose this should not play any role. This scenario has already been tested, and in the current setting it turned out to be very bad - some of the participants often fly out, sometimes it happens that only half of the participants hear the speaker. I sin on bugs in the flash client, but maybe something is wrong with the settings? After all, people use it.

2. One person has audio-video turned on, the rest only listen and sometimes ask questions in the chat or submit a request to turn on their audio-video. The number of participants is large, hundreds. Have not tried it yet, but I'm afraid that there will be problems on the current configuration. Has anyone come across such a scenario?

I would be grateful for any hints both on setting up the OM and on alternative solutions.

Answer 1

Also used OM. True, version 1.8 so far. When there were 8-9 people in the room, problems already started: sound interruptions, departures from the room of individual users, and everything like that you described. True, the server is weaker. But, as top showed, the load on the server was not high during the conference.
No matter what they tried to do. The channel was expanded, the tsiska was changed (all this was in the plans, it was done not only for the sake of conferences). In general, now we are testing a cloud service from Cisco Webex - we decided to abandon free solutions.
Although I'm still thinking of updating the version to try, or put BigBlueButton for testing.

Answer 2

Measure the channel. 15-20 people eat up to a gigabit, and therefore fall off.