Communications of the ACM
It's A Cloud, but Not As We Know It
A startup based in Great Neck, NY, is confident it can put a new form of cloud data storage into low Earth orbit.
Credit: Ozy Media Inc.
After India created a blizzard of lethal space debris by destroying one of its own satellites with a ground-launched anti-satellite missile earlier this year, low Earth orbit might not seem the best place to site a revolutionary new form of cloud data storage, one that requires a precision choreography of laser beams between many tens of satellites.
Yet LyteLoop, a startup based in Great Neck, NY, is confident it can succeed with just such a plan sometime in the mid 2020s—and even evade the dangers of space debris. For just over three years, the company, led by CEO Ohad Harlev and chairman and owner Daniel Damaghi, an industrialist, has been developing a patent-pending storage technique that turns conventional thinking about hyperscale cloud storage on its head.
"We were brainstorming about how to make data less hackable about three and a half years ago, when we wondered why companies had to have their data somewhere at all. Why can the data not simply travel around in a loop all the time? That way, it would make it a lot harder to find, catch, and hack," says Harlev.
For six months, Harlev recalls, they played with this idea of how to "store" data without giving it a physical location, deciding that ultra-high-capacity light-based solutions would be the answer, whether in optical fibers, vacuum tubes, cavities - or outer space. They then built a team that could realize such an idea, including CTO Paul McManamon, a photonics expert and former chief scientist for the U.S. Air Force's sensor directorate, and laser optics specialist Alan Willner of the University of Southern California, a former president of the Optics Society of America. They then drafted a patent application, which they filed in March 2017 (under LyteLoop's original name, LoneStar).
In a nutshell, the idea they are attempting to patent is that instead of storing, say, petabytes (thousands of terabytes) of data in the enormous, power-hungry, expensive-to-run server farms inside cloud data centers, they could instead store data by making it modulate—or piggyback on—endless laser beams that continuously loop between satellites orbiting the Earth. In other words, the data is never stored in the conventional sense; hence their technique's name, "storage in motion."
The idea is that data to be stored is uplinked via a radio frequency link to a master satellite, which then modulates a laser beam of a certain wavelength with that data. This is then transmitted to the next satellite, where it is reflected off mirrors on it to the next spacecraft, and so on until it circumnavigates the planet. Now and again, a laser on another satellite in the loop will be activated to regenerate the beam's signal strength (like the repeaters on telecoms systems). When new data needs to be added, it can be either added to an existing loop, if capacity allows, or a new loop can be established using different mirrors and lasers on different wavelengths on the same satellites.
Harlev thinks storage in motion could revolutionize the terrestrial cloud storage industry, which is becoming overly power hungry at a time when environmental and climate change issues make that impossible to justify on sustainability grounds. At the same time, once satellite-building and rocket launch costs have been ameliorated, a storage in motion cloud would have ongoing operational costs in the low millions of dollars per year, Harlev says, significantly less than the costs of cloud centers run by the likes of Amazon Web Services, Google, and Microsoft.
For that reason, says Harlev, a LyteLoop cloud could act as a cost-effective orbital backup for the datacenters of any of those computing industry behemoths. Other users could include banks wanting to have a backup data store that is well out of the reach of natural disasters, for example.
Still other LyteLoop users could include companies such as OneWeb and SpaceX, who are collectively planning to launch thousands of satellites (called megaconstellations) in a bid to provide broadband internet access from space, with LyteLoop's laser cloud perhaps providing them a vast local data cache.
Such services also hold lessons for LyteLoop. In late May, the first tranche of SpaceX's Starlink network went into orbit, with 60 tightly flatpacked satellites launched on a single Falcon 9 rocket. It didn't go well, however: that satellite swarm turned out to be visible from Earth, traversing the heavens like some kind of cosmic train. Over the next few days, astronomers became increasingly alarmed that when fully deployed, the 20,000 megaconstellation satellites planned by all vendors for low Earth orbit (LEO, altitudes below 2,000 km) could wreck the night sky for terrestrial astronomy.
However, Harlev says LyteLoop only plans around 100 satellites per storage loop, and these may be in Mid Earth Orbit (MEO, altitudes above 2,000 km) or Geostationary Earth Orbit (GEO, 36,700 km), where they will not be visible from Earth.
However, LyteLoop has yet to raise all the funding it needs to launch the project, and it has yet to determine precisely how many of the 500kg, 1-kilowatt satellites it will need in a loop, or into which orbits (LEO, MEO, or GEO) it will have to launch them.
The firm is nothing if not confident of its broad architecture, however, and Harlev expects orbital laser storage will strengthen data security in three ways. First, the data is moving around the planet at the speed of light, which will make it incredibly tough to intercept. Secondly, the data is "in space, so attackers will no longer be able to hack it from their mom's garage," he says. Finally, with data in motion using optical-based transmission, measures such as quantum key distribution and quantum cryptography can be employed to shore up the system. However, many quantum techniques are highly temperature-sensitive, and temperatures vary widely for a satellite in orbit, so if used, this technology will have a great bearing on the orbits LyteLoop chooses.
What is certain, however, is that the data in the loops will be protected against space debris, says Harlev, because each loop will be cloned: there will be two sets of satellites, each acting as backup to the other. If a satellite in a loop is hit by parts of a fragmenting rocket body or a destroyed satellite, the cloned loop will take over.
One observer in the computer science community who is used to investigating novel system architectures is guardedly impressed by the idea of data in motion. "This is a fascinating concept. If it could result in cheaper and more environmentally sound storage, then it is certainly worth exploring for the appropriate types of data," says Peter Bentley of University College London, a computer scientist specializing in radically new architectures such as computers that never crash and machine learning autopilots for airliners that learn by example.
Bentley is concerned about how LyteLoop can inject new data into its orbital loops at light speed and maintain low-enough error rates as it passes from satellite to satellite. Storage in motion, he says, might run the risk of becoming a digital version of the parlor game Telephone, in which players pass a message along a line of people to see how the message mutates as it is passed on. "Nature shows us that molecular storage [in DNA] works, but transmission of information will always, without exception, result in errors being propagated sooner or later; it's why evolution happens. So a data storage solution based on transmission will inevitably suffer the same fate. It's just a matter of time," he warns.
At IBM's research lab in Zurich, Switzerland, spokesman Chris Sciacca points out that there is little new under the sun: he notes that IBM was first to produce loop-based optical storage, with the development in 2006 of photonic buffer chips capable of storing as much data as 10% of the storage density of a floppy disk.
That such an outdated storage format as a floppy should have been cited is apt, however, as LyteLoop wants to render terrestrial cloud data centers just as redundant as that flexible 20th century medium.
Paul Marks is a technology journalist, writer, and editor based in London, U.K.
No entries found