Overview


Each vendor connection can be automatically monitored on 3 parameters - ACD (average call duration), ASR (average success rate) and PDD (post dial delay). You can choose to monitor one parameter or all of them.  


When one of the measured parameters passes your defined threshold then the connection can either be demoted to lowest priority, blocked, or no action (except for an alert email sent to the switch administrator).  When 3 of the measured parameters ("Quality Monitoring = "ACD, ASR, PDD") raises above your minimum threshold then the connection quality is considered good.

Quality measurement is done using the "Statistics Window" which consists of a configurable number of calls.


The system waits for the "Retry Interval, sec" and when the connection below the minimum threshold becomes unblocked (after N seconds) the system sends to it a batch of calls and based on the results of this batch of calls, the system makes a decision about connection quality.


The statistics window will be purged based on the count of successful calls in the retry batch if the connection's quality has changed from bad to good.
The new "Statistics window" will start gathering statistics from the first call in the "Retry Batch" that shows good quality.



Custom case 1:

Bad quality action: block

Retry batch value = 250 calls

Retry Interval = 1800 seconds


Connection is being used by 500 simultaneous sessions when the block action gets applied to any new call.

The retry interval starts from this moment, though the we do not kill the existing sessions.

After 250 calls have successfully finished, the system calculates the quality (ASR or ACD) and if the quality gets better the connection could be unblocked immediately otherwise the "block" record would be added to the connection_quality_stats and the connection stays in a block state.


Custom case 2:

Bad quality action: Make last in routing
Retry batch value = 250 calls
Retry Interval = 1800 seconds

Connection is being used by 500 simultaneous sessions when the block action gets applied to any new call.

The retry interval starts from this moment, though the we do not kill the existing sessions.

After 250 calls have successfully finished, the system calculates the quality (ASR or ACD) and if the quality gets better the connection could be set in its place immediately otherwise the "block" record would be added to the connection_quality_stats and the connection stays in its place.  Thought this connection can be last in routing it can terminate some part of the traffic (depending on the huntstop configuration and routing policy).  If this connection terminates 250 calls for a period of less then 1800 seconds, the system recalculates the quality of the connection using this batch even if the retry interval hasn't expired yet.



Connection Quality Recovery Algorithm


A bad quality connection is fully or partially disabled, so it is difficult to evaluate its current quality. To prevent indefinite blocking of a connection, the quality of a bad connections is periodically tested by sending a small percentage of calls to that connection. If the calls succeed, and the quality of that connection rises above the defined threshold  then the "Bad Quality Action" will be reverted.


Quality Monitoring Tunables


 The Edit Vendor Connection form contains the following options that you can set manually:

  • ACD Threshold - The Average Call Duration that is measured only for calls with non-zero duration, measured in seconds, When the ACD drops below this value then the connection is treated as bad, and the "Bad Quality Action" is executed
  • ASR Threshold - The Average Success Rate threshold is expressed as a percentage that describes the correlation between Total count of calls and non-zero duration calls. When the ASR percentage drops below this value the connection is treated as bad, and the "Bad Quality Action" is executed
  • PDD Threshold - The Average Post Dial Delays threshold is measured in seconds. When the PDD (in sec) jumps up above the specified value the connection is treated as bad, and the "Bad Quality Action" is executed
  • Statistics Window - The number of calls used to calculate the ACD and ASR metrics. ACD & ASR will be calculated on the last N calls made to the connection. The Statistics window is shifted when the new call comes. Example: if the window is 5 calls, either ASR or ACD is calculated for those 5 calls. Once the 6th call comes, the ASR or ACD is recalculated for the 2nd to the 6th calls, the same thing happens when the 7th call comes (the 3rd to the 7th calls are used to measure the new value)
  • Retry Interval - Number of seconds after which the switch will start sending a "Retry" calls to determine if the connection's health has improved from bad to good
  • Retry Batch - Number of calls to send as "tests" to test if the health of the connection has improved from bad to good
  • Bad Quality Action- When the chosen metric drops below the defined threshold, you can have one of the following actions take place:
    • Make Last In Routing - The Vendor connection will be dropped to lowest priority in the associated routing group(s). Retry calls will be sent after the 'Retry Interval' timer has elapsed, the connection will be temporarily returned to its normal place in routing. If the Retry calls succeed, and the quality metric rises above the threshold, the connection priority will be returned to its previous position in routing on a permanent basis, until the quality goes down again
    • Block - The Vendor connection will be blocked. Retry calls will be sent after the 'Retry Interval' timer has elapsed, the block will be removed temporary. If the Retry calls succeed, and the quality metric rises above the threshold, the block will be removed fully until the quality goes down again
    • Notify Only - No action will be taken, but the root user of the switch will be notified via email that is configured in his preferences
  • Notify On Status Change - This causes a messages to be sent to the e-mail address of the root customer on each change of the status of a connection. The messages are sent only on real not temporary status changes.