134 ServerIron ADX Global Server Load Balancing Guide
53-1002437-01
Smoothing mechanism for RTT measurements
1
This release introduces a new smoothing mechanism along with a proprietary smoothing algorithm
for GSLB RTT measurements to effectively deal with variances in RTT measurements. These
mechanisms allow you to define what is a very high or a very low value for an RTT sample on the
GSLB ServerIron ADX. If the new sample is in the acceptable range, GSLB ServerIron ADX will do a
smoothing similar to the one described above. If the value is much higher than current RTT value,
then GSLB ServerIron ADX will ignore this value a few times. If GSLB ServerIron ADX still sees this
large value after ignoring it for some time, then it will factor this value into existing RTT using an
additive increase. Similarly, if the value is much lower than current RTT, GSLB ServerIron ADX will
ignore it a few times. If GSLB ServerIron ADX still sees this small value after ignoring it for
sometime, then it will factor this value into the existing RTT value using a multiplicative decrease.
In the scenario described above, you can, for example, specify that anything more than 50% of the
existing value should be considered a very high RTT and should be ignored once. If RTT was 20ms
and the Site ServerIron ADX reported a new RTT of 1 second, then the GSLB ServerIron ADX will
ignore this value once. If Site ServerIron ADX continues to report an RTT of 1 second, then this will
be slowly factored into existing RTT value, using an additive increase.
Configuring enhanced RTT smoothing
To configure Enhanced RTT Smoothing, complete the following tasks.
1. Enable Enhanced RTT Smoothing. Refer to page 135.
2. Configure the parameters. Refer to page 136.

Parameters to smooth RTT variances

You can configure or modify the following parameters to customize enhanced RTT smoothing for
your network:
•Maximum deviation allowed: This parameter defines the maximum acceptable deviation for an
RTT sample. It defines what the GSLB ServerIron ADX should consider as a very high or a very
low RTT value.
For example, you may observe that the deviation between the old and new RTT value during a
spike in the RTT is typically 500% of the existing RTT value. You can use this information to
determine the optimal value for maximum deviation allowed. You can fine tune the value of
maximum deviation allowed which in turn will determine which RTT samples should be
considered as very high or very low as compared to the current RTT. The default value for this
parameter will be 400%.
•Maximum ignore count: This counter defines how many consecutive very high or very low new
RTT samples to ignore before factoring them into the existing RTT value.
This count specifies how many extremely high or extremely low RTT values to ignore. The
default is 3. If a spike in the RTT occurs once in a while, set this parameter to 1. If you do not
want to ignore any samples, set this parameter to 0.
•Normal ramp factor: This parameter defines the factor by which a RTT sample in the
acceptable range should be factored into the existing RTT value. This is typically around 10%.
•Ramp up factor: This parameter specifies the increments in which successively new high RTT
samples should be factored into the existing RTT value.