TCP checks from our US nodes attempt a three-way handshake to the target host and port, reporting connection time and success or failure. This tests actual firewall policy and port availability independently of ICMP. Common ports to test from US nodes include 80 and 443 for web, 25 and 587 for SMTP, 3306 or 5432 for database access verification, and custom application ports for game servers or API endpoints. TCP checks reflect the path that real application traffic takes, not a simplified ICMP echo.

The Miami, Dallas, and Kansas City nodes each sit on different upstream carriers with different BGP routing policies. A TCP connection from AS22295 in Kansas City may take a different physical path to a target in Europe than the same connection from AS202673 in Miami — and those paths may have different firewall exposure. Testing from all three US nodes verifies that no single upstream carrier is blacklisting the target port or applying asymmetric routing that breaks the return TCP path.

Failed TCP connections from US nodes when the port is known open from other regions typically indicate IP geoblocking, ASN-level firewall rules, or a misconfigured security group on cloud infrastructure. Cross-reference against our Canadian node (Montreal) and European nodes to determine whether the block is US-specific or tied to the specific AS. AWS security groups and GCP firewall rules are common sources of US node TCP failures — verify the allow rules include the probe IP ranges.

The US internet backbone is anchored in Ashburn, Virginia, where Equinix operates the largest concentration of interconnected networks in the country. Over 500 networks peer at the Equinix campus in Ashburn, including AT&T (AS7018), Comcast (AS7922), Lumen/CenturyLink (AS3356), Cogent (AS174), and NTT (AS2914). The density of peering there means that a packet originating in Miami or Dallas often transits through Ashburn before exiting to Europe, making it the effective default gateway for US-to-Europe traffic regardless of where the origin server sits.

Our US nodes span three cities across the South, Central, and Midwest regions. Miami runs on AS202673 (Ohz Digital SL) and is the southernmost probe — useful for measuring connectivity relevant to Latin American networks and Caribbean-facing services. Dallas runs on AS201129 (Linceris International Cloud) at Dallas DC, positioned at the intersection of south-central US routes with good proximity to both Equinix Dallas peering and central US backbone routes. Kansas City runs on AS22295 (Advin Services LLC / Tierhive) in the Midwest, providing a third vantage point with direct access to central US carrier paths.

Reference RTTs from these nodes under normal load: Miami to London is approximately 120 ms, Miami to São Paulo around 85 ms. Dallas to Frankfurt typically runs 130–135 ms via transatlantic submarine cables landing on the US East Coast. Kansas City to New York is roughly 35 ms, and Kansas City to Los Angeles is in the 40–45 ms range. These figures vary by carrier — Cogent and Lumen have different peering strategies at Ashburn, which produces measurably different latency for the same destination depending on which AS the probe exits through.

Key US transit providers reachable from all three nodes include AT&T (AS7018), Comcast Business (AS7922), Lumen (AS3356), Cogent (AS174), and NTT America (AS2914). These five carriers collectively carry the majority of US internet traffic and each maintains presence in Dallas, Kansas City, and Miami in addition to Ashburn. At the IX level, DE-CIX New York and NYIIX serve the Northeast corridor. Equinix Dallas and CoreSite serve south-central US. Miami is served by the NAP of the Americas facility, a major Latin America-facing interconnect point operated by Equinix.

Running checks across all three US nodes simultaneously provides geographic diversity within a single country that matters. A server hosted on AWS us-east-1 will show different RTTs from Miami (via southeast paths), Dallas (via south-central), and Kansas City (via central US backbone). A CDN with US PoPs should show low single-digit or sub-10ms results from all three. If one node shows significantly higher latency than the others, the routing from that city or the serving carrier is suboptimal for that node's upstream transit mix.