← ROOTBEER

R&E Anycast Prefix Explorer

Anycast prefixes routed in R&E Measured through

Anycast prefixes

Search

R&E non-R&E median RTT > slower in R&E < faster in R&E

Loading…

Overview

AS labels

Ribbons

Font size 10px

Consolidate ASes off

Export

Legend

Speed-of-light ÷ 1.0

Data sources: CAIDA Ark, CAIDA AS-to-Organization (AS names), Twente Anycast, ISI hitlist, and RouteViews.

Snapshot info

Loading…

We collect active anycast measurements daily and combine them with these reference datasets, captured as of the measurement run:

CAIDA Ark — vantage-point platform for the ping/traceroute measurements.
CAIDA AS-to-Organization — AS numbers to network names and holders.
Twente LaCeS — census of anycast prefixes and instance locations.
ISI hitlist — responsive IPv4 targets to probe within each prefix.
RouteViews — BGP data for prefix origins and AS paths.

How to read the AS-path graph

This view is a Sankey diagram of the AS-level traceroute paths observed for the selected anycast prefix. It answers: from which vantage-point ASes did probes start, which transit ASes did they traverse, where did paths merge near the origin, and whether R&E and non-R&E vantage points saw different route families.

RTT CDF Summary

The compact chart above the Sankey is a CDF of measured RTTs for the selected prefix. The x-axis is RTT in milliseconds. The y-axis is the cumulative percent of each class of vantage points at or below that RTT.

The blue curve is R&E vantage points; the orange curve is non-R&E vantage points.
Each curve is normalized within its own class. That means the shape is comparable even when there are many more non-R&E vantage points than R&E vantage points.
A curve that rises farther to the left means that class has more vantage points reaching the anycast service at lower RTT.
The dashed vertical markers show the class medians: 50% of that class is at or below the marker and 50% is above it.

The Reading Direction

Read the graph left→right. Vantage-point ASes are on the left, transit ASes are in the middle, and the prefix's origin AS or origins sit on the right. The columns are based on AS-hop distance to an origin: ASes directly before an origin appear next to it, while earlier ASes sit farther left. The diagram is not a geographic map, a BGP policy graph, or a claim that one path is canonical; it is a layout of measured traceroute paths.

Bars: ASes

Each vertical bar is one autonomous system. A bar appears only when that AS participates in at least one observed AS-to-AS hop.
Bar height reflects the amount of observed path traffic through that AS. Very large route families are visually softened so smaller but meaningful alternatives remain visible; hover the bar for exact R&E and non-R&E trace counts.
The origin AS uses the teal accent colour. Vantage-point source ASes have a heavier outline. Other ASes are blue for R&E membership and orange for non-R&E membership when that classification is known; when classification is missing, the graph falls back to the dominant observed trace class through the AS.
Labels show the AS number. The AS labels setting in ⚙ Graph settings toggles network names in the labels; the full name and holder are still available on hover. The same dialog sets the label font size.
The Consolidate ASes slider in ⚙ Graph settings strips vantage-point-level detail while leaving the overall flow to the origin intact. Any AS carrying less than the chosen share of total vantage-point flow is spliced out, and paths that then match are merged into one ribbon — collapsing many individual vantage points into the shared trunk. Origins and the ASes feeding them directly are always kept.

Ribbons: Measured Path Families

By default, the graph uses collapsed ribbons: path families that share the same displayed AS-to-AS hop and class are aggregated into one wider band. This keeps large prefixes readable and makes the graph behave like a conventional Sankey overview.

The Ribbons setting in ⚙ Graph settings switches to Full detail, which draws each measured path family as its own band — useful for detailed inspection but can make large prefixes visually busy.
R&E path families are placed in a soft upper lane and non-R&E path families in a soft lower lane. Shared ASes can bridge the lanes when both classes traverse them.
Ribbon width is proportional to the number of traceroutes in the displayed aggregate or path family, with the same visual softening used for bars. Hover a ribbon for the trace count and, when aggregated, the number of path families in the band.

What Counts as Passing Through an AS

Ribbons only connect adjacent columns. If an observed path jumps over a column, the renderer threads that ribbon through an invisible lane between bars instead of drawing it behind a bar it did not traverse. So a bar carries only ribbons for traceroutes that actually include that AS. Nearby ribbons in the gaps are passing by, not passing through.

Full AS Paths

The graph shows every observed AS hop in each measured path up to the first listed origin AS. The origin is treated as the destination for this view, so traceroute ASNs that appear after an origin are not drawn as outgoing paths beyond the service. Ribbons and their tooltips use these displayed AS sequences.

Colour and RTT Inflation

Node colour describes AS membership: R&E, non-R&E, or origin.
Ribbon colour describes latency, not membership. The green→yellow→red scale compares a ribbon's end-to-end RTT with the best observed RTT for this prefix.
Green means the path family is close to the best-served vantage point. Yellow and red indicate increasing RTT inflation. Inflation can come from a longer route, a distant anycast catchment, geography, queueing, or measurement noise; the colour identifies where to investigate, not the cause by itself.
For separate ribbons, the colour uses the median RTT of the traces in that exact AS-path family.
For collapsed ribbons, several path families that share the same displayed AS-to-AS hop and class are bundled into one band. The colour uses a trace-count weighted average of the bundled path-family median RTTs. This makes high-volume path families contribute more to the aggregate colour than low-volume path families.
The collapsed-ribbon tooltip reports the same aggregate RTT value used for colour, plus the number of path families in the bundle. Switch to Full detail when you need the exact median RTT and colour for one underlying family.
The legend maps the colour scale from the best observed RTT to the 90th percentile RTT for this prefix, which keeps a single extreme outlier from making everything else look identical.

R&E vs Non-R&E Toggles

The R&E and non-R&E buttons in the header show or hide entire classes of path families. Hidden ribbons still reserve their layout space, so the remaining class stays in the same position and you can compare before/after without the diagram rearranging itself.

Useful Things to Look For

Early split, late merge: R&E and non-R&E probes use different upstreams but converge before the origin. This often points to access-network or transit differences rather than a different origin.
Different penultimate ASes: path families enter the origin from different providers or peerings. This is often the clearest sign of different anycast catchments.
Shared AS, different colour ribbons: both classes traverse the same AS sequence, but one class has higher RTT. That suggests latency is not explained by the AS path alone.
Red path family with high width: many vantage points are consistently seeing inflated RTT through the same route family, making it a stronger candidate for operational follow-up.

Interacting

Hover an AS for its AS number, name, holder, role, membership, and R&E/non-R&E trace totals.
Hover a ribbon for the AS-to-AS hop, class, trace count, and full AS path family. The ribbon colour encodes relative RTT inflation across the prefix (see above).
Click an AS to focus it. Ribbons that are not part of a traceroute through that AS dim. Click the AS again, or click the background, to clear the focus.
Click a ribbon to list the vantage-point traceroutes crossing that hop, sorted by RTT. Pick a vantage point from that list to open its full measurement detail: RTT, ASN sequence, hostnames, and hop IPs where display is allowed.

Origins and MOAS

The right edge is anchored by the origin AS or origin ASes for the prefix. If the prefix is a MOAS prefix, more than one origin-coloured bar may appear. Paths terminating at any listed origin are treated as reaching an origin for layout purposes.

Limits of the View

The graph is AS-level. It hides router-level hops unless you open an individual measurement.
Traceroute paths are measurements, not complete BGP route advertisements. Missing hops, unresolved ASNs, and traceroute artifacts can affect what appears.
RTT is end-to-end for the vantage point, so a red ribbon does not prove that the AS-to-AS hop itself is slow.
The layout optimizes readability of measured path families. Vertical position is meaningful for grouping and comparison, but it is not geography or local preference.

Sharing and Export

The page URL updates with the selected prefix, selected view, AS-name setting, ribbon mode, and focused AS, so you can share the same graph state. The Export buttons download the graph as SVG, PNG, or PDF and include the surrounding context: prefix name, Internet2 path, anycast instance locations, caption, and data-source credits.