Observations on a Europe-bound IPv6 path

2026-11-21

Sampled a path between a home connection and a European VPS over a day. Nothing exotic, but the pattern was clearer than I expected:

windowICMP RTTsingle-flow TCPnotes
evening peak~140-150msfluctuates, retransmits climb5-8% sampled loss on short probes
after midnight~138-145msstable ceilingloss nearly vanishes

Takeaways:

Moral: before blaming your access line, sample the same path at 2am.

Back to notes