Phoromatic Tracker Strides Forward

[phoronix]

Phoronix: Phoromatic Tracker Strides Forward

Following in the success of the Phoronix Test Suite, last month we launched Phoromatic as a remote test management system targeted for enterprise users of the Phoronix Test Suite that allows the automatic scheduling of tests, remote installation of new tests, and the management of multiple test systems all through an intuitive, easy-to-use web interface. Whether you are looking to build a test farm or just benchmark systems around the world, Phoromatic can turn this otherwise taxing work into a really easy process with turn-key deployment capabilities. As an extension of Phoromatic, we then wrote Phoromatic Tracker that is designed to track any software component (either on a timed or per-commit basis) and automatically execute a set of tests each time around all in an autonomous matter and then pump the data back to the Phoromatic server and showcase it on the Phoromatic Tracker interface...

http://www.phoronix.com/vr.php?view=ODA2MA

[sabriah]

This looks sweet!

[sabriah]

Phoronix: Phoromatic Tracker Strides Forward

Oh, I notice you use the geometric mean. What is the rationale for that?

* The geometric mean is typically used for processes which show exponential growth, often NP-hard issues. It is not very sensitive to outliers.

* The harmonic mean is typically used for rates, such as MB/sec. It is not very sensitive to outliers.

* The median is very good at finding trends and is the least sensitive to outliers and crap data.

* The arithmetic mean is the most sensitive to outliers and random data.

As you want to find the outliers as indicators of a regression, and not ignore them, I guess the traditional arithmetic mean may after all be the most relevant for you. Any takers?

[Michael]

Oh, I notice you use the geometric mean. What is the rationale for that?

* The geometric mean is typically used for processes which show exponential growth, often NP-hard issues. It is not very sensitive to outliers.

* The harmonic mean is typically used for rates, such as MB/sec. It is not very sensitive to outliers.

* The median is very good at finding trends and is the least sensitive to outliers and crap data.

* The arithmetic mean is the most sensitive to outliers and random data.

As you want to find the outliers as indicators of a regression, and not ignore them, I guess the traditional arithmetic mean may after all be the most relevant for you. Any takers?

Thanks for the feedback, sabriah. Statistics is not my forte and geometric mean was just one of the composite graphs mentioned by Matthew. I just added in support for harmonic means to Phoromatic Tracker. Arithmetic mean will likely go in too.

[Michael]

More fun stuff @ http://www.phoromatic.com/kernel-tracker.php

And a few announcements are pending.