jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From abalanonline <...@git.apache.org>
Subject [GitHub] jmeter issue #296: Bug 61078 - Percentile calculation error
Date Wed, 10 May 2017 04:12:10 GMT
Github user abalanonline commented on the issue:

    https://github.com/apache/jmeter/pull/296
  
    Hello @pmouawad ,
    Different estimation types used by Apache Math explained [here](https://en.wikipedia.org/wiki/Quantile#Estimating_quantiles_from_a_sample).
    The use case for nearest rank method is simple. If the 90% percentile is 1200 ms than
that means that 90% of tests take no more than 1200 ms. Same for 95% and 99%. Very informative
value for JMeter. It is implemented in R_1 and R_2.
    The purpose of interpolated method is to eliminate the discontinuity and to get data useful
for calculation in different linear functions. Difference is in the min and max values. If
we treat them only as starting and ending points in the diagram then C=1 variant (R_7) is
used. Otherwise if the min and max are the full value experiment results and deserve their
own space on a diagram then C=0 variant (LEGACY, R_6) is considered.
    Intermediate variant C=1/2 (R_5, R_8, R_9) is fine too because it is balanced and does
not distort the "angle" of a diagram.
    The bug affects most of the libraries because of a cognitive inertia of developers. The
ceiling function is very rarely used and it is psychologically easier to use round instead.
Also round function give correct median for odd number of experiments. Seeing that they calm
down and believe that everything is calculating correctly.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message