Regional Details Dashboard

Detailed performance distributions for the cross-product of selected client ISPs and M-Lab servers near an anchor metro. Plot lines that lie together mean uniform service; lines that spread apart reveal servers that perform better or worse for some users.

The Anchor Metro defines the region: it preselects the M-Lab servers within the chosen radius and ranks the client ISPs by traffic in that metro. A combined PDF + CDF chart is then shown for every selected server crossed with every selected client ISP, with test counts in the legend. See the paper or slides for how differences in these plots reveal asymmetric mid-path routing that can adversely affect users.

Selectors

• Anchor Metro — geographic centre used to preselect servers and rank client ISPs. Changing it re-populates both Servers and Client ISPs.
• Radius (km) — how far from the anchor to include additional M-Lab servers.
• Servers — select M-Lab sites within the radius; defaults to all.
• Client Rows — how many top-ranked ISPs to pre-process.
• Client ISPs — select a subset of Client ISPs to plot; defaults to all selected.
• Extra rows — appears when the selected servers span multiple metros. Client ISPs are ranked independently per metro, so an ISP's rank can differ between them; Extra rows pulls in additional ranked ISPs so cross-metro comparisons aren't truncated.
• Metrics — which metrics to plot: MinRTT (log and linear), MeanThroughputMbps, LossRate.
• Table — whether to show the summary-statistics table (none / Summary / Verbose), and Table Field — which metric it reports.
• Method — cached (default), live, or live-DS16. Date pickers appear only for the non-cached methods.

Reading the charts

Each chart row is one client ISP; each chart is one metric, with one line per M-Lab server site. The figure is split vertically — PDF (probability density) in the bottom half, CDF (cumulative distribution) in the top half — sharing the same x-axis and a consistent color per site. Legend labels include the total test count.

Cached data

With method=cached the notebook uses pre-computed histograms; the actual data date range is shown below the selectors after clicking Run / Refresh.