ParallelRequestsAggregator generates a line plot that shows the number
of parallel requests among the plotted events.
This plot has two issues:
1. It only considers events that are plotted. Events that occur later,
but were started within the plotted time frame are not considered.
2. For performance reasons we are only plotting points when a value
changed. This leads to diagonal lines.
Unfortunately the datetime picker does not support seconds. But it is
one of the few that support date and time and are flexible enough to
be used with VueJS.
1. yRange is set to 0-120 minutes. This makes the gallery plots easier
to compare.
2. Set groupBy to pod/method/build. That is what I use the most and I
think this will help in most cases.
- split the 'sortby' select field into two fields
- sort by average
- legend shows plotted and total values in the date range
- removed InlineDataSeries, because it was not used anymore
Sometimes it is useful to specify the certain y-axis range. For example
when you are only interested in the values that take longer than a
threshold. Or when you want to exclude some outliers. When you want to
compare plots in a gallery, it is very handy when all plots have the
same data-area.
Eventually we want to only support what is now called aggregate, but
not have to implement different plot types. So instead of supporting
percentile plots for dashboards I removed them. You can still get
percentile plots together with the scatter plot.
Scaling big plots to small thumbnails results in bad images that barely
show any details.
We solve this by calling gnuplot a second time to generate the
thumbnails. They don't have any labels and are rendered in the required
size, so that not scaling is necessary.
Thumbnails have to be requested explicitly, because it can be expensive
to compute them.
The last improvement of memory usage introduced a performance
regression. The ingestion performance dropped by 50%-80%, because
for every inserted entry the Tags were created inefficient.
For each entry we executed two calls to System.nanoTime().
The resulting numbers aren't very reliable and calling nanoTime
that often (160k - 300k per second) is quite expensive.
