FreeSWITCH configuration and deployment

switchio relies on some basic FreeSWITCH configuration steps in order to enable remote control via the ESL inbound method. Most importantly, the ESL configuration file must be modified to listen on a known socket of choice and a park-only extension must be added to FreeSWITCH’s XML dialplan. switchio comes packaged with an example park only dialplan which you can copy-paste into your existing server(s).

Event Socket

In order for switchio to talk to FreeSWITCH you must enable ESL to listen on all IP addrs at port 8021. This can configured by simply making the following change to the ${FS_CONF_ROOT}/conf/autoload_configs/event_socket.conf.xml configuration file:

-- <param name="listen-ip" value="127.0.0.1"/>
++ <param name="listen-ip" value="::"/>

Depending on your FS version, additional acl configuration may be required.

Park only dialplan

An XML dialplan extension which places all inbound sessions into the park state should be added to all target FreeSWITCH servers you wish to control with switchio. An example context (switchiodp.xml) is included in the conf directory of the source code. If using this file you can enable switchio to control all calls received by a particular FreeSWITCH SIP profile by setting the "switchio" context.

As an example you can modify FreeSWITCH’s default external profile found at ${FS_CONF_ROOT}/conf/sip_profiles/external.xml:

<!-- Contents of  -->
-- <param name="context" value="public"/>
++ <param name="context" value="switchio"/>

Note

You can also add a park extension to your existing dialplan such that only a subset of calls relinquish control to switchio (especially useful if you’d like to test on an extant production system).

Configuring software under test

For (stress) testing, the system under test should be configured to route calls back to the originating FreeSWITCH (cluster) such that the originator hosts both the caller and callee user agents (potentially using the same SIP profile):

FreeSWITCH cluster                   Target test network or
                                     device

--------------   outbound sessions   ---------------------
| Originator | --------------------> | Device under test |
|            | <-------------------- |   (in loopback)   |
--------------   inbound sessions    ---------------------

This allows switchio to perform call tracking (associate outbound with inbound SIP sessions) and thus assume full control of call flow as well as measure signalling latency and other teletraffic metrics.

Example proxy dialplan

If your system to test is simply another FreeSWITCH instance then it is highly recommended to use a “proxy” dialplan to route SIP sessions back to the originator (caller):

<!-- Proxy Dialplan - forward calls to requested destination -->
<condition field="${sip_req_uri}" expression="^(.+)$">
    <action application="bridge" data="sofia/${sofia_profile_name}/${sip_req_uri}"/>
</condition>

Note

This could have alternatively be implemented as a switchio app.

Configuring FreeSWITCH for stress testing

Before attempting to stress test FreeSWITCH itself be sure you’ve read the performance and dialplans sections of the wiki.

You’ll typically want to raise the max-sessions and sessions-per-second parameters in autoload_configs/switch.conf.xml:

<param name="max-sessions" value="20000"/>
<!-- Max channels to create per second -->
<param name="sessions-per-second" value="1000"/>

This prevents FreeSWITCH from rejecting calls at high loads. However, if your intention is to see how FreeSWITCH behaves at those parameters limits, you can always set values that suit those purposes.

In order to reduce load due to logging it’s recommended you reduce your core logging level. This is also done in autoload_configs/switch.conf.xml:

<!-- Default Global Log Level - value is one of debug,info,notice,warning,err,crit,alert -->
<param name="loglevel" value="warning"/>

You will also probably want to raise the file descriptor count.

Note

You have to run ulimit in the same shell where you start a FreeSWITCH process.