Matrix Cable System

Based on 39 RIPE Atlas measurements from GeoCables monitoring infrastructure, March–April 2026.

The Matrix Cable System — operated by Matrix NAP Info, a regional Indonesian carrier — is a 1,055-kilometre submarine cable with three landing points: Batam in Indonesia's Riau Archipelago, Jakarta on the north coast of Java, and Changi South in Singapore. The cable was placed in service in 2008 and has been operational for eighteen years. As regional Southeast Asian cables go, Matrix is unusual in two ways. First, it is a private cable — owned by a single mid-sized Indonesian carrier rather than by a consortium of national incumbents. Second, of its three landings, two are inside Indonesia and one in Singapore, which makes it primarily an Indonesian backbone cable with a Singapore tail rather than an international interconnect.

What makes Matrix interesting from a measurement perspective is something that no single number can capture. Looking at the cable's average round-trip alone tells one story. Looking at it broken down by direction tells a completely different one. The cable does not have one personality. It has two, and they live on opposite ends of the same physical fibre.

Across 18 RIPE Atlas measurements from Jakarta to Changi South in Singapore, the average round-trip is 14.41 ms, the minimum is 14.21 ms, and the standard deviation is 0.24 ms. Across 21 measurements in the opposite direction — Changi South to Jakarta — the average is 114.68 ms, the minimum is 69.91 ms, the maximum is 288.86 ms, and the standard deviation is 45.01 ms. Both endpoints are the same. Both directions traverse, nominally, the same submarine cable. One side delivers nearly perfect performance. The other side shows the kind of variance we usually see on cables in much less mature peering environments.

Two cables, one wet plant

The physics floor for the Jakarta–Singapore segment of Matrix is 10.33 ms. The Jakarta-side measurement of 14.21 ms minimum sits at 1.376× the floor — close to the cleanest result possible on a 1,055-km route, with the small overhead attributable to the shore-side equipment on each end and the trace through the carrier's domestic network before reaching the cable terminal. A standard deviation of 0.24 ms means almost no variance over the measurement window. Whatever path the Indonesian-originated traffic takes to Singapore, that path is stable, well-engineered, and clearly going through the Matrix cable as designed.

The Singapore-side measurements describe a different system. The minimum of 69.91 ms is roughly seven times the floor — five times worse than the reverse direction. The average climbs to 114.68 ms. The maximum reaches 288.86 ms. The standard deviation of 45 ms means observations are scattered widely; the path is not a single stable route but a varying mixture of routes, some of which evidently avoid the Matrix cable entirely. Singapore-originated traffic to Jakarta is being load-balanced or dynamically reselected across multiple alternative paths, only some of which go through the cable.

The asymmetry is not a property of the cable itself. Submarine fibre is symmetric by design; light travels at the same speed in both directions of a fibre pair, and the wet plant has no mechanism for introducing direction-dependent latency at this scale. The asymmetry is introduced in BGP. Each side of the connection independently decides how to route its outbound traffic, and those decisions are governed by transit cost, peering agreement structure, and the AS-path length seen for each potential next-hop. The Jakarta-side carrier evidently treats the Matrix cable as the preferred outbound path to Singapore. The Singapore-side carriers treat it as one of several possible paths and route accordingly.

Why direction-as-quality is a real pattern

This kind of direction-dependent asymmetry is not unique to Matrix; we see variants of the same shape on many regional cables in Southeast Asia. BRCS (Batam-Rengit Cable System) shows extreme asymmetry between forward and reverse, with both directions far above the floor — neither side uses the cable preferentially. DMCS (Dumai-Melaka Cable System) shows extreme variance dominating both directions — multiple path choices changing over time. Matrix sits between these two patterns: one direction is firmly committed to the cable, the other direction is not.

What this commonly reflects, when seen on Indonesian regional cables, is the structure of the local market. Indonesian carriers exchanging traffic with Singapore typically prefer to use submarine capacity they own or have committed contracts on — reaching Singapore IXPs adds transit cost even when Singapore IXPs are physically close. Singapore-originated traffic to Indonesian destinations, by contrast, does not have the same asymmetric incentive: from Singapore, Indonesian destinations are one of many South Asian options, and content delivery networks routinely cache content in Singapore that ends up serving Indonesian users via paths Singapore carriers chose for their own reasons.

The result is a cable like Matrix where one direction carries a steady operationally-disciplined flow and the other direction is a load-balanced mix of best-effort routes. Both sides are measuring the same physical cable. They produce different numbers because they live in different routing systems.

Where Matrix sits in the regional set

Matrix's Indonesian-direction floor multiplier of 1.376× is among the cleanest we measure on regional Southeast Asian cables. Gondwana-1 between New Caledonia and Sydney also sits at 1.081× — essentially at the floor — for the same reason: a single committed path with minimal alternative-route competition. ARCOS-1 in the Caribbean measures at 0.613× — below the floor — because the route chords across the basin via Florida hubs.

The Singapore-direction multiplier of 6.78× tells the opposite story: a route that is using the cable only some of the time. Comparing the two directions of Matrix against each other is itself instructive — the same wet plant, the same fibre pair, the same physical end-to-end length, with one side measuring at 14 ms and the other at 70-289 ms. This is the floor multiplier as a measurement of routing commitment, not of cable quality.

What direction-asymmetric cables tell us about the regional market

The wider story Matrix illustrates is structural to Southeast Asia's submarine cable economy. Indonesian carriers historically committed substantial capital to building outbound capacity — Matrix is one example, B2JS (JaKa2LaDeMa) is another, DMCS is a third. From the Indonesian side, those investments translate into routing preferences: ID-domestic and ID-outbound traffic uses owned capacity. From the Singapore side, the same cables are options among many, and Singapore's depth as a peering ecosystem means most ID-bound traffic ends up passing through Singapore IXPs first regardless of which submarine cable eventually carries the bits.

Matrix's two-personality measurement signature is, in this sense, an honest reading of that asymmetric market. A cable owned by a single Indonesian operator is committed-to in one direction and treated as one option among many in the other. The 14 ms direction is what disciplined routing looks like on this 1,055-km route. The 114 ms direction is what BGP looks like when the route is one of several. Both are measuring the same fibre. The fibre is fine. The asymmetry lives in the network around it.