AWS Region-to-Region Latency Matrix

Every commercial AWS region pinging every other. 35 × 35 = 1,225 cells. Hover for exact values, click a header to lock a row/column highlight.

The diagonal (dark cells) is intra-region latency — always ~1 ms. Everything else is inter-region.
Color scale runs green → yellow → red across the observed range for the selected percentile. Same color in both p50 and p99 views means the link is stable; a big red jump when you switch to p99 means jitter.
Rows sorted by geographical group: EU → MEA → NAM → SAM → APAC. This makes the "same-continent" green blocks pop out visually.
The matrix is usually near-symmetric (A→B and B→A route through similar AWS backbone paths), but not always. A few pairs — especially anything from remote regions to eusc-de-east-1 (European Sovereign Cloud) — show large asymmetries: e.g. ap-southeast-2 → eusc-de-east-1 is 614 ms but the reverse is ~310 ms. That's a real property of the sovereign-cloud egress path, not a measurement bug. Marked cells with ^! exceed the color-scale clamp (P95 of all pairs); hover to see the actual value.
Sub-millisecond precision is spurious — treat values as ±3 ms.
Bahrain (me-south-1) shows suspiciously stable values across percentiles — sparse sampling in the source, not a real property of the link.

Reading the matrix