Re: adaptive threshold significance level

From: Dragutin Jastrebic <orahawk_at_gmail.com>
Date: Tue, 4 Nov 2014 20:29:32 +0100
Message-ID: <CANGCQw=3obKErFCUGn2Brb7j-ee5opfbOHBzp0JU0QwfGjermQ_at_mail.gmail.com>



 =>
By reading the above statement it seems t me that it is saying for warning level it will rise an warning alert when 0.05 percent of observations are above the threshold and raise a critical alert when it is 0.01 percent. Should not it be the opposite? When a higher number of observation is higher than the threshold a critical alert should be raised? <=

No, it is not the opposite, this is exactly the way it is supposed to work, since the Significance Level adaptive treshold is used to capture seldom, unusal values,

For exemple, with OEM 12c you have a metric "Average number of active sessions", with the
following defaults:

Very high level (0.99) for a warning treshold Extreme level (0.9999) for a critical treshold.

The value that has 1 occurence in 100 will trigger a warning alert, but the value that has 1 occurence in 10000 will trigger a critical alert

So more unusual the value is, it is estimated as more critical.

BTW, it seems that OEM 12c with its quick configuration, is mostly suggesting Significance Level
for OLTP and Percentage of Maximum for DW systems. This make sense: we are more interested to capture seldom values in OLTP that in DW database

Beside Oracle documentation, for more information about the adaptive tresholds, you have a good explanation on Asktom=>

https://asktom.oracle.com/pls/apex/f?p=100:11:0::::P11_QUESTION_ID:1525205200346930663

Also, I have written a little article about this topic on my blog, this is strictly nothing new, but there are 2 pictures that can help the global understanding of adaptive tresholds

http://jastrebicdragutin.wordpress.com/2014/11/03/adaptive-tresholds/

HTH
Dragutin

>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Tue Nov 04 2014 - 20:29:32 CET

Original text of this message