Leo Speed: When
Milliseconds Are Worth $Millions
October 14, 2020 by Carlos Placido | NSR
“Remember that time is
money” wrote Benjamin Franklin in a 1748 book.
The concept of opportunity cost believed
to trace back to ancient Greek philosophers but
popularized by one of the founding fathers of
the United States, may nowhere be more critical
to business than in the realms of high-frequency
trading (HFT). Planned LEO and VLEO (Very Low
Earth Orbit) satellite constellations cannot
provision data rates as high as fiber cables
when connecting major cities. Yet, speed is also
about link delay and a subset of these
high-throughput satcom initiatives will be able
to offer lower latency than fiber across long
distances, a situation that could foster wider
interest in satellite constellations by
financial players.
A large portion of trading
today is happening between computer servers with
sophisticated algorithms that trade hundreds of
thousands of times a second, seeking to
repeatedly make small amounts of profits over
tiny periods of time. Stock price is subject to
constant change, leading to arbitrage
opportunities that HFT algorithms seek to
quickly capitalize on via speed of execution.
Whoever accesses information first has an
enormous advantage in speculative markets so, in
the age of cloud computing and artificial
intelligence, promptly leveraging data can
provide a solid competitive edge. Disregarding
the controversies around HFT including
market-crash risks derived from relying too
heavily on computers that play robot wars and
make snap decisions on behalf of humans, it can
be plausibly argued that a competitive edge of
10 milliseconds is worth millions of dollars for
hedge funds and other financial trading
stakeholders.
Can LEOs indeed provide
lower latency than fiber?
Let us benchmark satellite
and fiber link delay between two of world’s most
connected financial districts: New York City and
London. It is a fact that light travels faster
in vacuum so based on NSR’s Non-GEO
Constellations Analysis Toolkit, even the
shortest-path, “great circle” submarine fiber
connection between New York and London has 28%
higher propagation delay than a LEO
constellation with satellites orbiting at an
altitude of 550 kilometers (or lower), provided
satellites have inter-satellite links (ISL).
A theoretical shortest-path
fiber connection between NYC and London is 5,577
km (3,465 miles) long (assuming Earth radius of
6,378 km). Over such minimum distance, fiber
cannot deliver round-trip latency below 55
milliseconds due to fiber’s light refraction
index, which essentially makes light travel
around 40% slower than in vacuum. Indeed,
recorded latency metrics (PINGs) between NYC and
London across commercial submarine fiber cables
consistently show a round-trip delay floor of 70
milliseconds. Commercial fiber cables have
multiple hops and network nodes so rarely run
along “great-circle”,
shortest paths between two end points.
Nevertheless, there is a fiber-optics network
implementation with ultralow latency purposely
built to target financial players that achieves
the lowest possible delay, close to the
theoretical limits: GTT’s
acquired Hibernia Subsea cable route provides a
round-trip latency of just under 59 milliseconds
between New York’s NY4 and London’s LD4 exchange
data centers.
A LEO satellite
constellation with orbits at an altitude of 550
km can be 12 milliseconds faster than fiber
optics (round trip propagation latency). This is
a theoretical benchmark that ignores jitter
(satellites are on the move) and processing
delay added by active equipment (such as
demodulating/remodulating signals) but note that
if the satellite network can be configured in
cut-through mode across ISLs to minimize
processing, the delay introduced by the two end
point modems can be negligible. As an example,
using Datum’s performance calculator (actual
satellite modem implementation) a DVB-S2X
Carrier at 155 Mbps with 64K FEC Block
introduces only 0.3 ms end-to-end.
Thus, even multiple link legs (with
different ModCods) could result in a negligible
FEC processing latency.
Alternate Satellite
Topologies
Interestingly, when running
alternate satellite and hybrid satellite-fiber
scenarios on NSR’s Non-GEO Constellations
Analysis Toolkit, it can be found even without
ISLs, a 550 km or lower altitude constellation
shell may still beat fiber optics’ shortest-path
propagation latency by a small margin – assuming
no terrestrial interconnect segment. This
“ground bounce” scenario, as coined by Elon
Musk, assumes use of intermediate “relay
stations” (either plain Gateways used as
“pivots” or user terminals with peer-type
interconnect capabilities). Note that the
NY-London example may not the best example for
such particular scenario given difficulties in
deploying relay stations in the middle of the
Atlantic Ocean but it paints the picture.
Even classic “Hub & Spoke”
satcom network topologies employing
terrestrially interconnected gateway stations at
one or two ends could -in theory- provide lower
propagation delay than fiber; albeit by small
margins for the 550 km altitude case. Given
delay variations introduced by satellites on the
move, these topologies may not provide much
advantage in practice, so ISLs will be key.
Bottom Line
The idea of pushing the
limits by leveraging alternate technologies to
achieve the lowest possible link latency is not
new. This is one reason why microwave links
connect the Chicago and NY stock exchanges but
LEOs and VLEOs could extend such ultra-low
latency capabilities to distant exchange sites.
There is, interestingly, a past satellite
attempt: LeoSat, which in 2019 needed to go back
to the drawing board to reduce the
constellation’s price tag, saw the opportunity
and announced in 2016 that it had entered into
an agreement with an (undisclosed) globally
operating financial trading company with offices
worldwide.