Using a clustered chart to compare distributions

[gaglenn]

I have an issue where I am trying to use a clustered bar chart to provide A/B distribution comparisions.  The problem I am encoutering is when calculating the percentage of grand total, it is using the global grand total as a denominator, rather than to total for each individual dataset.  This is causing the larger dataset to swamp-out the smaller one, and make the comparison not useful.  Is there a setting or option to segregate the category totals for the clustered data sets?

To illustrate, the graph below shows the primary distribution:

The next graph shows the secondary latency distribution:

The clustered chart would ideally be a composite of these two distribution graphs.  However, when combined the %GT ioCount appears to cover the ioCount for both latency categories, giving a graph as follows:

This is causing the green category to be dominant, and the black category to be statistically insignificant.  Is there an option that would normalize these data sets better?

[amitchandak]

@gaglenn , In such cases you should prefer combo visuals and use line for the second measure. Like Line clustered column chart. But that will show vertical bar

Or try Log axis

[gaglenn]

Thank you for the suggestion.  In this case, I just opted to go for two side-by-size bar graphs, as it seemed the clustered graph was not going to work the way I wanted it to.