Asterisk scaling

Home » Asterisk Users » Asterisk scaling
Asterisk Users 2 Comments

Sorry for the top post, this is from my phone.

What you need to look at are the following:

Is it going to be just one mp3 stream that is shared across all users (I.e everyone hears the same thing at the same time), or is it 1000 separate mp3 streams (everyone always starts at the beginning of whatever they are hearing).

Are you going to have reliable timing generation on an EC2 instance, since IAX streams and music on hold playback will sound bad if the timing isn’t good.

Will you have sufficient bandwidth allocated to you for that many simultaneous calls?

Is there going to be any codec transcoding going on? Can you generate your streams in the preferred codec, instead of mp3?

I think if you’re just using one stream spread across all the callers, you’ll have much better performance from the system as a whole. You may want to look at the quality differences between a SIP trunk and an IAX trunk as well.


2 thoughts on - Asterisk scaling

  • It’s a shared stream. When testing now, new listeners doesn’t spawn new
    mpg123 processes.

    We are using the zaptel and ztdummy kernel module, and we haven’t
    noticed any problems with the audio quality yet. Should we be worried
    about this when the load gets higher?

    Good point. We will have to do some calculation and research on what EC2
    offers here.

    The source is an icecast server streaming mp3. I haven’t figured out a
    way to get around that. But from what I understand its just one
    reencoding for all the listeners.

    I had a talk with our IAX2 trunk provider and they told me that we could
    expect better performance from a SIP trunk. They also had a limit on
    2000 channels, so we may have to look for another trunk.

    Are there any tools or services to simulate a lot of IAX2 or SIP users
    that you can recommend? How do you test how many users an asterisk
    system can handle?

    Thank you for taking the time to reply.