FMS 3.5 cpu usage

[ffmpeg]

due to customer requirement for his app we have to use the old 3.5 version, the server is an Intel L5420 (8 cores), 16 giga RAM with centos 6.8 -64 bit. it goes up to 700 -800 video chat users, when it goes over that the app restarts although not the FMS. Monitoring while at 500 users, the cpu usage was 65%.

The same app works fine on a X5660 Intel  processor,with more or less the same system and RAM.

Is there anything we can do on settings side so to use the 5420 which is much cheaper then the 5660?

thanks

[vayvanne]

800 users on 16 GB is less than 20 MB per user. my bet is RAM not CPU. have you tried to monitor swapping?

[ffmpeg]

Hi vayvanne,

I don't think so, both servers have 16 giga of RAM, and monitoring on the fms admin or through server root the RAM usage is little,  surprisingly. What we notice is the CPU usage which on the L5420 goes way up.

[vayvanne]

are the FMS configs on both servers identical?

[ffmpeg]

how can I check that? what are the files usually altered for the fms configuration?

[vayvanne]

conf\server.xml

conf\_defaultRoot_\_defaultVHost_>VHost.xml

conf\_defaultRoot_\_defaultVHost_>Application.xml

[ffmpeg]

wow thanks. what do I look for on this files? is it to optimize the fms? what values should I be looking for?

[vayvanne]

check for differences on the failing and working servers... if there is such, try one by one from the working server)

[ffmpeg]

thank you vayvanne, I will try that.

[ffmpeg]

the only difference was on the file

conf\_defaultRoot_\_defaultVHost_>Application.xml

where the original fms which is the failing server has

  <!-- This section contains information about configuring the script engine. -->

    <JSEngine>

        <!-- This specifies the max size (Kb.) the runtime can grow to before -->

        <!-- garbage collection is performed.                                 -->

        <RuntimeSize>1024</RuntimeSize>

on the working server the 1024 value has been replaced with 30720

would that be relevant?