Latency in LEO Satellites vs. Terrestrial Fiber

By | July 7, 2021

Latency is a key differentiating parameter for new communication networks such as LEO satellite constellations and 5G. The interest in latency has increased as the value proposition for these networks hinges on enabling latency-dependent applications. This got some competing technology options nervous. For instance, LEO satellites provide as much as 20x round trip delay advantage over GEO satellites. On the other hand, LEO satellites provide a more nuanced advantage over terrestrial fiber which depends on distance. We researched this aspect of LEO satellites in collaboration with Angola Cables (see here) to understand the potential market dynamics between the satellite and subsea fiber ecosystems.

Independently, Aizaz Chaudhry at Carleton University was doing a deep technical analysis to quantify the latency advantage of LEO satellites over terrestrial fiber using sophisticated satellite constellation simulator among other tools. Here, I summarize some of the key aspects related to data transport over LEO satellites that Aizaz and I arrived at.

 Quantifying the Value of Latency

The latency advantage of LEO satellites accounts for a good part of their value proposition. The business case for LEO satellites is related to the users’ willingness to pay for latency. However, latency impacts user experience differently according to the type of traffic. Low latency could enable new applications that would otherwise be impossible, or it would have no difference on the performance of other applications (e.g. email). The non-uniform value of latency means the price it commands could be different depending on the use case and the user. The bet on LEO satellites is in part a bet that users value latency, as otherwise they can obtain service from GEO satellites albeit at 20x the latency.

Addicted on Latency

One market segment that craves latency and is willing to pay for it, is high-frequency trading (HFT). In a report published in January, 2020, the UK Financial Conduct Authority (FCA) estimated the global size of latency-arbitrage races at $5 billion per year. Other highlights include:

  1. The average FTSE 100 symbol has 537 latency-arbitrage races per day; or about 1 race/minute/symbol.
  2. The winner beats the first loser by 5-10 microseconds.
  3. About 22% of daily trading volume for the FTSE 100 index is in latency-arbitrage races.

Another often quoted metric – which is rather dated now – by Information Week estimates the value of a 1 msec advantage to a major brokerage firm at $100 million.

The FCA report cites that to achieve low latency, some market centers are connected with dedicated straight-line microwave links and fiber optical cables that minimize the distance, such as GTT Express submarine cable (formerly Hibernia Express). Even the transport equipment leverages hardware implementation using FPGAs and SoCs as opposed to general purpose processors to avoid software-based solutions which have slower data processing and response time.

LEO satellites may ultimately not be a solution for HFT, but the application highlights the importance of latency to a specific sector. In our analysis, we compare the latency performance of LEO satellites to terrestrial fiber (land or subsea) because potentially one could quantify a value for a part of the overall LEO satellite value proposition.

Space vs. Terrestrial Latency

Propagation in free space is 50% faster than in fiber optical cables. This gives satellites an advantage over terrestrial fiber after exceeding a distance that accounts for the altitude of the satellite. For a constellation such as SpaceX at an orbit of 550 km, this breakeven distance is around 2,700 km. A constellation such as Telesat Lightspeed at 1,325 km, the breakeven distance is around 9,000 km. Both these figures are approximations since satellites move and the actual distances change accordingly. To link any two points on earth, LEO satellites would require to implement inter-satellite links (ISL) to route traffic from one satellite to another.

RTT Latency comparison between terrestrial/subsea fiber and LEO satellites.
Approximate shortest path RTT Latency comparison between terrestrial/subsea fiber and LEO satellites. [Source: Xona Partners]

Aizaz’s paper (see here) describes the details of how ISL could work to transport data between two points on earth. I would refer the reader to review it as it clearly describes some of the challenges that could rise from routing satellites in different orbital planes. Using a constellation simulator, Aizaz calculated the difference in latency between the following cities for a LEO constellation at 550 km orbital altitude:

Latency (ms)Improvement
Fiber
(terrestrial/submarine)
Satellitems%
New York–Dublin25.0720.07519.94
Sao Paulo–London46.5736.649.9321.32
Toronto–Sydney76.2958.3417.9523.53

The improvement in latency represents the shortest path between two cities and are specific to the satellites’ orbital altitude and planes. One could add to this baseline the delays resulting from routing in the satellites to compare between different satellite constellations (which we hope to do in the future; routing within satellites has low delay, on the order of few hundreds of microseconds). Doing so would account for the tradeoff between lower altitude and higher number of routing satellites vs. higher altitude and lower number of routing satellites.

Space vs. Terrestrial Cost of Capacity

Terrestrial fiber, including submarine, provides Tbps-capacity that cannot be matched by satellites. Moreover, the cost of capacity in terrestrial fiber is 1000x less than that in space. As a result, terrestrial fiber is ideal to transport large amounts of data, whereas satellites are ideal to transport latency sensitive data over long distances.

Cost of capacity for LEO satellites and submarine fiber cables.
Cost of capacity for LEO satellites and submarine fiber cables. [Source: Xona Partners & Angola Cables]

The latency and capacity tradeoffs make terrestrial and satellite transport complementary technologies. Terrestrial fiber wins in most applications, especially where it’s readily available. However, satellite transport wins where the amount of data is not comparatively large, but where latency is highly valued, as in the case of HFT along with a host of new application.

Concluding Thoughts

Not all satellite constellations feature ISL. For instance, SpaceX first generation satellites are not equipped with ISL. On the other hand, constellations such as Telesat Lightspeed will have ISL from the start. ISL increases the cost of the satellites to account for additional equipment, weight and power requirements. However, this additional increase is well worth it, especially to save on the number of ground stations and to enable new applications such as maritime and aviation connectivity. See here if interested in how different constellations compare and contrast.

Finally, if interested in the latency requirements of application, check out the presentation on “Competing for the Edge: Analysis of Competitive Dynamics in Edge Computing.”