Almaty to Tokyo: 877 ms via London and Singapore — a 21,000-Kilometre Internet Detour
Based on RIPE Atlas measurements from a probe in Almaty, Kazakhstan, targeting endpoints in Tokyo, Japan. Measurement window: April 2026.
The minimum round-trip we observe between Almaty and Tokyo is 877 milliseconds. The great-circle distance between the two cities is approximately 5,400 kilometres. At the speed of light in optical fibre, a packet making that journey end-to-end would arrive in about 53 milliseconds round-trip. We measure sixteen times that. The traceroute reveals why.
The packet leaves Almaty on local fibre operated by Signal Telecom and JSC Transtelecom — Kazakhstan's two principal national carriers — and reaches Astana within 5 milliseconds. From Astana, it does not turn east toward China, nor north toward Russia and onward to the Pacific. It turns west. The next significant hop is London, on AS9002 (RETN Limited), at 17 milliseconds. Then 28 milliseconds. Then, in a single jump, the round-trip leaps from 28 milliseconds in London to 251 milliseconds in Singapore. That single jump — over 222 additional milliseconds — represents the submarine cable transit from Western Europe across the Indian Ocean to Southeast Asia. From Singapore the packet continues to Osaka, then to Tokyo, ending its journey at 877 milliseconds in the Tokyo metropolitan area, on AS59105 (Home NOC Operators Group).
This is six countries, four named cities, and approximately 21,000 kilometres of fibre to reach a city that is geographically 5,400 kilometres away. The detour is structural: it reflects the absence of high-capacity submarine cable infrastructure across the Asian heartland and the resulting BGP convergence on London as the practical Central Asian peering hub.
Why Almaty traffic goes to London
Central Asia — Kazakhstan, Uzbekistan, Turkmenistan, Tajikistan, Kyrgyzstan — is one of the largest landmasses on the planet without a single submarine cable landing. Every byte of internet traffic to or from the region must cross a national border by terrestrial fibre. The available terrestrial routes have been shaped by geopolitics as much as by geography: the Russian fibre corridor north toward Moscow has been the historic default, but it has been progressively de-preferred since the early 2020s by carriers seeking to avoid Russian transit; the Chinese corridor east is constrained by the Great Firewall and limited peering arrangements; the Iranian corridor south is constrained by US sanctions on Iranian transit infrastructure; and the Caucasus corridor west, reaching the Black Sea coast through Azerbaijan and Georgia, has limited capacity and crosses several geopolitically sensitive borders.
What remains, for the bulk of Central Asian commercial internet traffic, is a long terrestrial route across the Caspian basin and through eastern Europe — typically Kazakhstan to Russia or to Azerbaijan to Turkey to the Black Sea — and from there onto the European fibre backbone toward Frankfurt, Amsterdam, and London. London, specifically, is where Central Asian carriers find the densest peering ecosystem and the cheapest IP transit. Turkmenistan's entire international internet has historically been carried by three foreign IP transit providers, all of which terminate the Turkmen side of the connection somewhere in continental Europe. The Azerbaijan-Kazakhstan corridor across the Caspian Sea remains underserved by submarine fibre despite repeated planning efforts. Kazakhstan as the largest of the Central Asian nations is the best-connected of the bunch, but its outbound routes still funnel toward European peering hubs by default.
RETN, the carrier on whose AS9002 our packet first reaches London, is a Pan-Eurasian fibre operator that runs terrestrial backbone across Russia and into Europe, with Asian extensions through Mongolia and direct submarine connectivity from London onward. For Central Asian carriers, RETN is one of the standard transit options: a single backbone that can deliver Kazakh traffic from Astana to London in under 20 milliseconds, where it can then peer with the rest of the global internet.
The 222-millisecond submarine cable jump
Between hop 8 (London, 28 ms) and hop 10 (Singapore, 251 ms) lies the longest single segment of the entire path. That 222-millisecond increase represents about 22,000 kilometres of round-trip fibre — roughly the length of one of the SEA-ME-WE family cables connecting Europe to Asia via the Mediterranean and the Suez Crossing.
SEA-ME-WE-4, in service since 2005, is one of the principal cables carrying this traffic. It lands at Marseille in France, transits the Mediterranean to Alexandria on the Egyptian coast, crosses Egypt by terrestrial backhaul through Suez, then resumes underwater across the Red Sea, the Indian Ocean, the Strait of Malacca, and finally to Tuas in Singapore. The total path from London to Singapore via this corridor is around 11,000 kilometres of submarine fibre plus a few hundred kilometres of terrestrial fibre across Egypt and a few hundred more on the European side. The 251-millisecond Singapore arrival is consistent with this geometry: it is what an honest single-traversal of this corridor produces.
Other cables on the same corridor — SEA-ME-WE-3, SEA-ME-WE-5, IMEWE, EIG, AAE-1 — produce broadly similar latency budgets. MENA Cable, an alternate that GBI built specifically to provide redundancy on this corridor, is also part of this set. The Almaty-to-Tokyo packet may have used any of these cables on the London-Singapore leg; the traceroute does not tell us which one. What it tells us is that one of them was used, and that the resulting latency is exactly what the corridor's topology dictates.
Singapore to Tokyo: the second submarine jump
From Singapore, the packet jumps to Osaka in Japan, then continues to Tokyo and the Mejiro suburb where the test endpoint sits. The Singapore-to-Japan corridor is one of the densest in Asia and is served by SJC (Southeast Asia-Japan Cable), JUPITER, FASTER, and several other systems. The traceroute hop counts climb rapidly through Japanese carrier networks — AS2518 BIGLOBE for one hop, then AS59105 Home NOC Operators Group for several more — and the maximum measured latency on individual hops reaches 928 milliseconds before settling at 877 milliseconds for the final Tokyo destination. The variance in the per-hop times suggests that the Japanese terrestrial network was congested or that intermediate routers were buffering packets at the time of measurement; the 877-millisecond figure represents the steady-state observation, not the worst case.
SJC, in particular, has a Singapore landing at Tuas and a Japanese landing at Chikura on the Pacific coast. Japan's Pacific cable cluster lands several Asian-origin cables at this same coastal area before backhauling to the Tokyo metropolitan data centres. The final 250-300 milliseconds of round-trip latency between Singapore and the Tokyo measurement endpoint are consistent with this corridor's geometry plus terrestrial backhaul on the Japanese side.
Why this won't change soon
The natural alternative to this 21,000-kilometre detour would be a direct terrestrial fibre corridor from Central Asia to East Asia: Almaty to Ürümqi in Xinjiang, then to Lanzhou, Beijing, and onward to Japan via either northern China or via Korea. Such a route would be perhaps 7,000 kilometres total, and the resulting Almaty-Tokyo round-trip could plausibly be in the 70-80 millisecond range — about ten times faster than the current 877-millisecond detour.
The barriers to building this corridor are entirely non-technical. The Chinese network operates under different peering and transit rules than the rest of the global internet, and traffic routed through Chinese AS numbers can encounter packet loss, latency spikes, or routing inconsistencies attributable to deliberate filtering at the network level. Carriers that use Chinese transit risk having their traffic affected by the same controls that affect Chinese domestic users, even when the destination is outside China. This makes Chinese transit unattractive as a default routing option for international commercial traffic, even when the geographic shortcut is enormous. Central Asian carriers, faced with this trade-off, route west to London instead — accepting an extra 800 milliseconds of round-trip latency in exchange for predictable, controlled, peered transit.
A second alternative would be direct submarine cables landing on the Caspian Sea coast of Iran or Turkmenistan, with terrestrial backhaul across the Caucasus to Black Sea cables and onward to Asia by a southern route. Iran's submarine cable footprint is currently small, and US sanctions complicate the participation of Western consortium owners in any new cable touching Iranian shores. Without an Iranian or Turkmen submarine landing, the southern alternative remains hypothetical.
The third alternative — direct submarine cable from Central Asia to East Asia via the Arctic and the Bering Strait — has been proposed in various forms (Polar Express on the Russian Arctic coast, the proposed Far North Fiber connecting Europe to East Asia via the Arctic) but none of these are operational on a large scale. For now, the London detour is the practical default and is likely to remain so for years.
The 877-millisecond reality
What we measure on the Almaty-Tokyo route — 877 milliseconds, six countries, two submarine cable jumps, and a routing decision shaped by geopolitical constraints rather than by the geographic minimum path — is an honest snapshot of how the Central Asian internet actually reaches East Asia in 2026. The figure is not anomalous. It is the equilibrium that BGP, peering economics, and regional infrastructure availability have settled on. Until any of those three change, Almaty-to-Tokyo traffic will continue to take its 21,000-kilometre journey through London, the Mediterranean, the Suez Crossing, the Indian Ocean, the Strait of Malacca, Singapore, and the Pacific — to reach a city only 5,400 kilometres away.
