Threshold sets the upper threshold value for calculating network monitoring statistics created by an IP SLAs operation.
Timeout sets the amount of time an IP SLAs operation waits for a response from its request packet. Both are in milliseconds.
Frequency sets the rate at which a specified IP SLAs operation repeats, in seconds. You have it set to 3 seconds.
If you configure Timeout > Threshold, you can see the round-trip-time and if it exceeds Threshold, it is a failure.
If you configure Threshold > Timeout, if the round-trip-time exceeds timeout but less than Threshold, it won't show the
round-trip-time (as nothing happens till threshold) but it is still a failure.
**Please grade this post if you find it useful.
Thanks for your reply. Could you explan bit more on threshold.
Does that mean threshold value should be half of the timeout value configured?
threshold 250 -- Is it a one way time
timeout 500 -- Is it a round trip time
frequency 3 -- Will it check the peer every 3 seconds once
Could you please let us know the best practise configuration or sample config as we are facing interminant sla down alert.
I want to understand more about threshold. for example if a threshold value is 200ms,
what it does? Does it mean system wait for 200ms for SLA re-computation.
There a common practice for setting those 3 values: (Frequency) > (Timeout) > (Threshold). For network intensive SLA such as UDP jitter, a minimum 'Frequency' of 60 seconds is advised, and that's because of the load and the overhead that they put on the network.
'Threshold' doesn't have to be set at any specific value, it is totally up to you and your vision/plans for the network's service and quality level. Although it must be set smaller than the value that's set for the 'Timeout'.
In another word; 'Time out' = no service, unreachable, dead. Where 'Threshold' is a value that defines a warning level before your service completely goes out. When the upper threshold value is reached, your device will execute the task that you have configured it to do, e.g. send a Syslog message or trigger a SNMP Trap, etc.
plz Rate if it helped.