Tbilisi to Yemen: 790 ms via Frankfurt and Starlink — How War-Disrupted Aden Reaches the Internet
Based on RIPE Atlas measurements from a probe in Tbilisi, Georgia, targeting endpoints in Aden, Yemen. Measurement window: April 2026.
The minimum round-trip we observe between Tbilisi and Aden is 790 milliseconds. The traceroute reveals an unusual route: Tbilisi exits Georgia through Sofia and Vienna, reaches Frankfurt at hop 10 in 60 milliseconds, and then — at hop 14 — hands off to AS14593, Space Exploration Technologies Corporation. The next visible hop is Aden, on the same AS14593, at 790 milliseconds. The packet reaches Yemen not through any submarine cable, but through Starlink, the SpaceX-operated low-Earth-orbit satellite network.
This routing is not what one would expect from a connectivity-by-cable model of the internet. Yemen is on the southern edge of the Arabian Peninsula, just across the Bab-el-Mandeb strait from the Horn of Africa, and historically has been served by submarine cables landing on the Red Sea coast at Salif and Hodeida and on the Gulf of Aden coast at Aden itself. The cables — FALCON (Reliance), FEA (Flag Europe Asia), AAE-1, and others — are part of the same Red Sea corridor that SEA-ME-WE-4 and MENA Cable use. In a healthy state of regional connectivity, traffic from Tbilisi to Yemen would route via European peering hubs to Marseille or another Mediterranean landing, then through the Suez crossing, down the Red Sea by submarine cable, and into Yemen via one of the coastal landings.
The Starlink hop tells us something has changed.
What AS14593 Starlink does in this path
Space Exploration Technologies Corporation, operating Autonomous System 14593, is the network ASN for Starlink, the satellite-internet service that began commercial operation in 2021 and now covers most of the world's land area through a constellation of low-Earth-orbit satellites in approximately 550-kilometre orbits. Starlink's architecture has the user terminal communicate with overhead satellites, which then route data either directly to a ground station near the destination (if one exists) or laterally through inter-satellite laser links to a different ground station closer to the destination, where the traffic re-enters the terrestrial fibre network.
For most Starlink users, the latency profile is in the 30-60 millisecond range — a clean LEO transit between user terminal, satellite, ground station, and onward fibre. The 790-millisecond observation on the Tbilisi-to-Aden path is well outside this normal envelope. The implication is that the path is not a single Starlink LEO transit; it almost certainly involves additional segments that compound the latency. The most plausible structure is: Tbilisi-side ground egress through European fibre to Frankfurt, where the Starlink ground gateway routes the packet up to a LEO satellite, the satellite either routes laterally through inter-satellite links toward the Yemen region or performs another ground-station-to-satellite hop, and the destination Starlink terminal in Aden receives the packet. The accumulated latency reflects Starlink network internal routing plus any congestion or capacity constraints on the Yemen side of the system.
The presence of AS14593 in the destination of a Yemeni IP address tells us that the destination endpoint is using a Starlink dish for its connection, not a terrestrial ISP or a submarine-cable-backed home connection. In practical terms, this means the endpoint we are measuring is one of the growing number of Yemeni users — government, commercial, individual — who have adopted Starlink as their primary or backup internet path.
Why Yemen needs satellite-based internet now
Yemen's submarine cable connectivity has been catastrophically disrupted since the early 2020s. A series of submarine cable cuts in the southern Red Sea between 2024 and 2025 took down multiple Asia-Europe cables simultaneously, including SEA-ME-WE-5, IMEWE, AAE-1, EIG, and others, with the cuts widely attributed to the conflict in the area and the threat from anchor strikes by vessels manoeuvring under wartime conditions. The repair operations have been complicated by the security situation; submarine cable repair ships require predictable safe access to the cut location, and the southern Red Sea has not been a safe operating environment for civilian commercial ships for several years.
For Yemen specifically, the consequences have been acute. The country's terrestrial fibre infrastructure depends on the Red Sea cable landings at Hodeida (Houthi-controlled) and Aden (Yemeni government-held) as its connection to the rest of the world. With multiple cables in the corridor inoperable for months at a time, and with the political division of the country making cross-front terrestrial backhaul fragile, Yemen has been operating with significantly degraded international connectivity. The downstream rerouting effects on Oman and other Red Sea-adjacent countries have been documented in our earlier coverage of the cable cuts.
Starlink became commercially available in Yemen in 2024, with the SpaceX terminal certification and ground gateway provisioning that allowed Yemeni users to subscribe to the service through several international resellers. For users in Aden — the government-held southern port city — Starlink provides an alternative to the submarine-cable-dependent terrestrial network: the satellite connection bypasses the wet plant entirely and reaches the rest of the internet through SpaceX's ground gateways in Europe and elsewhere. The Tbilisi-to-Aden path we measure is one of the consequences of that arrangement: the destination endpoint is Starlink-connected, the source endpoint reaches it through normal European fibre transit, and the bridging segment is the Starlink network itself.
Why Tbilisi traffic exits through Frankfurt
The Tbilisi-side portion of the route is unremarkable by Caucasus internet standards. Georgian carriers have historically used several European peering hubs as their primary international transit points: Frankfurt, Vienna, and London are all reachable in 50-70 milliseconds from Tbilisi over standard Pan-Eurasian fibre backbones. AS1299 Arelion (formerly Telia Carrier) is a common upstream provider for Georgian carriers, and the path we observe — Tbilisi to Sofia to Vienna to Frankfurt — is a textbook Arelion routing through eastern European peering hubs into the central European peering core.
From Frankfurt, the packet would traditionally continue through European fibre to Marseille, across the Mediterranean by submarine cable, through the Suez Crossing, and down the Red Sea to a Yemeni landing. SEA-ME-WE-4 is one of the cables that, in a healthy operating state, would carry exactly this kind of traffic. With the Red Sea corridor disrupted, the BGP path that would have used SEA-ME-WE-4 or its sister cables is unavailable, and the next-best path advertised by the Yemeni endpoint's network is the Starlink AS14593 path. The handoff from terrestrial Arelion fibre at Frankfurt to Starlink at the same Frankfurt gateway is an explicit BGP routing decision: Yemen via Starlink is reachable; Yemen via submarine cable, at the time of measurement, is not.
This is the same kind of structural answer we see in Iran's submarine cable footprint being shaped by sanctions, in Turkmenistan's reliance on three foreign IP transit providers, or in Jerusalem to Tanzania traffic transiting via Djibouti — the routing reflects not just where cables exist but also which paths are commercially and politically usable at the time of measurement. For Yemen in 2026, the politically and commercially usable path is via Starlink.
The 790-millisecond breakdown
Decomposed by segment, the 790 milliseconds we measure are approximately: 60 ms of terrestrial fibre transit from Tbilisi to Frankfurt, plus an additional 730 ms of Starlink network traversal from Frankfurt gateway to Aden user terminal. The Starlink portion is dominant by an order of magnitude, which is unusual for a Starlink connection. A typical user-to-user Starlink path with both ends in well-served regions would measure in the 60-100 ms range; the Tbilisi-to-Aden number suggests that the Yemeni-side terminal is being reached through a longer-than-direct Starlink network path, possibly because the Starlink ground gateway nearest to Aden is not optimal for the given route, or because the inter-satellite laser link routing chose a longer hop chain at the time of measurement. Starlink's internal routing decisions are not directly observable to external probe traffic; we can only see the latency consequence.
For users of the connection, the 790-millisecond round-trip is operationally significant. Standard interactive web traffic is acceptable but feels noticeably slow; voice and video communication is at the upper boundary of usability; and any latency-sensitive application — financial trading, real-time gaming, certain types of remote work — is impractical. The connection is, in practical terms, a fallback option rather than a first-choice path. But it is functional, and that functionality is what allows Aden's users to remain connected to the global internet during the period of submarine cable disruption.
Yemen's connectivity in transition
The Tbilisi-to-Aden 790-millisecond Starlink path is a snapshot of how Yemen's internet works in 2026, not a permanent architectural feature. As the Red Sea cables are repaired or replaced, Yemen's terrestrial connectivity will improve and Starlink's role will diminish. Yemen has historically not been a major submarine cable nexus, but it sits at one of the world's most strategic chokepoints, and the cables that pass through Bab-el-Mandeb on their way between the Indian Ocean and the Mediterranean are critical infrastructure for the global internet. Reconstruction of that infrastructure is a multi-year project that depends on the political and security situation as much as on the engineering capacity of the cable repair industry.
What we measure on the Tbilisi-Aden path — 790 milliseconds, the appearance of Starlink AS14593 in the destination ASN, the absence of any submarine cable in the trace — is an honest reading of Yemen's substituted connectivity in mid-2026. The country is online; it just gets there by a different physical route than the cable-system maps would suggest, and the cost of that substitution is paid in latency. We will continue to track the path. When submarine cable repair eventually shifts the routing back to the Red Sea corridor, that change will be observable as a sudden drop in latency on this exact pair of endpoints — at which point Yemen's wartime substitution will have ended.
