• Versatile
• Efficient
• Lightweight
• Modern
• Adaptable
• Versatile

Network Monitoring

Here’s a clear picture of the US↔Europe connectivity and latencies for the past week (May 27 – June 3, 2026):

Active Probes

(Several other probes exist but are disconnected or interrupted: Frankfurt, Google Cloud EU, OVH France, etc.)

Average One-Way Latency (last week) — NYC → Europe

US West Coast → Europe

Key Takeaways

• Latency is solid and consistent across all six transatlantic paths, with no major spikes or outages.
• New York → UK is the fastest at ~39 ms (shortest undersea cable route).
• US West Coast adds ~30–35 ms over NYC for the same European destinations.
• One notable spike: NYC→Helsinki had a single hour averaging ~232 ms with a P95 of ~557 ms (around May 28 06:00 UTC), suggesting a brief routing flap.

Spike Drill-Down: NYC ↔ Helsinki, May 27 ~06:00 UTC

The probe alternates between two measurement streams ~60s apart (different source ports → different ECMP hashes):

Stream B stayed at a rock-solid ~53 ms the entire time. Only Stream A was hit.

The spike was symmetric (both directions)

Both directions spiked simultaneously on Stream A, ruling out one-sided sender bufferbloat.

Verdict: Transient buffer saturation on a single ECMP path. The alternating pattern with only one stream affected is the smoking gun — one specific intermediate router was experiencing severe queue buildup (bufferbloat) from a traffic microburst. The incident resolved itself within ~2 minutes without any routing changes. Zero BGP announcements/withdrawals were recorded for either prefix during the window.

Here’s the full picture from the four traceroutes.

Hop counts (TTL depth, intermediary routers)

Average across all four directions: ~16 intermediary routers, but with very large asymmetry.

EU ↔ US East (NYC) — ~40–45 ms

Nuremberg → NYC uses Core-Backbone (AS33891) — a very direct 14-hop path:

Hetzner NBG → core-backbone.com nbg40
  → ae7-2093.nyk10.core-backbone.com
    → DigitalOcean NYC

NYC → Nuremberg uses NTT (AS2914) — 17 hops, routing via London or Paris before Frankfurt:

DigitalOcean NYC → NTT Newark
  → NTT London/Paris
    → NTT Frankfurt
      → Hetzner NBG

EU ↔ US West — ~82 ms

Nuremberg → us-west-1 uses Telia Carrier (AS1299) — 28 hops, the longest path by far:

Hetzner NBG → Telia Paris → Telia Ashburn (VA)
  → Telia Chicago/Atlanta
    → Telia LA/Palo Alto
      → AWS

us-west-1 → Nuremberg uses NTT (AS2914) — only 15 hops:

AWS → NTT San Jose (snjsca04)
  → NTT Ashburn (asbnva02)
    → NTT Frankfurt (frnkge13)
      → Hetzner NBG

Key Takeaways

• The asymmetry is entirely driven by ISP peering choices, not physical cable distance.
• The most extreme imbalance: 26 hops outbound vs. 13 hops inbound on the US West path — Telia’s US interior routing is much chattier than NTT’s.
• The NTT path (used by both US probes returning to Europe) consistently routes via London or Paris before Frankfurt, explaining the ~5 ms penalty on the NYC→EU direction vs. EU→NYC.