Logarithmic Scales to the Rescue — Librato Blog

Logarithmic Scales to the Rescue

Different Metrics, Different Distributions 

Some of the underlying systems that generate the measurements we use in our charts may not be normally distributed. The peculiarities of these distributions will make traditional render approaches ineffective.

The problem is that the wider range nature of these distributions causes certain values to dominate the plot. This problem is magnified when multiple streams are rendered within the same plot.

Alternative Scales

To help alleviate this problem, we can use logarithmic scales for our axes. With them, the transformation from the input domain to the visual representation will follow a nonlinear (logarithmic) transformation. This will boost the lower values of our dataset and help us to extract more value from the plot.

In the example image, we have the same stream rendered with linear (left) and logarithmic scales (right). You can see how the blue series dominate the graph and make it difficult to read and extract value from the green and orange series. Enabling logarithmic scales boost the lower values. Now the green and orange series are easier to perceive, helping us extract more value from our data.

Now you can enable logarithmic y-axis in your line charts. Click in the chart dropdown and select chart attributes. You’ll see a new option to enable this feature:

We hope this feature will help you to better understand certain types of your metrics. As always, any feedback is welcome: support@librato.com


Team Librato