Opened 9 years ago

Closed 7 years ago

#2564 closed enhancement (wontfix)

Visualize Torperf data as quantiles of moving windows

Reported by: karsten Owned by:
Priority: Low Milestone:
Component: Metrics/Website Version:
Severity: Keywords:
Cc: Actual Points:
Parent ID: #2769 Points: 6
Reviewer: Sponsor:

Description

Our current Torperf graphs show the daily median and quartiles of download times. This is fine for plotting intervals of a few weeks or more. But sometimes we want to plot Torperf data of a few days or even a single day (#2476). In these cases we might want quantiles of shorter time periods or, possibly even better, of moving windows. We might replace the graphs on the website with these new graphs, but more importantly we should have the code available for making custom graphs. Related to #2519 and #2563.

Child Tickets

Change History (5)

comment:1 Changed 9 years ago by karsten

We should also look into ggplot2's stat_quantile() some more. Maybe we can make it smarter than calculating a linear fit for all data points. Or maybe there's another function in ggplot2 that does what we want.

comment:2 Changed 9 years ago by karsten

Component: MetricsTorperf
Owner: karsten deleted
Points: 6
Priority: normalmajor
Status: newassigned

Setting priority to major, because we really need a better visualization of short experiments.

comment:3 Changed 9 years ago by mikeperry

Parent ID: #2769

comment:4 Changed 8 years ago by karsten

Component: TorperfMetrics Website
Priority: majorminor

The Metrics Website code aggregates Torperf data right now. Changing component to Metrics Website. Also, similar to #2519, this feature would be nice to have, but won't happen anytime soon. Reducing priority to minor.

When analyzing the effect of some performance change, one can always look at the raw Torperf data instead of watching how the graphs on the metrics website change or do not change.

comment:5 Changed 7 years ago by karsten

Resolution: wontfix
Status: assignedclosed

No, this won't happen at all. The metrics website code is complex enough with daily aggregated data. Having more detail based on moving windows would kill it.

Note: See TracTickets for help on using tickets.