back to article Space upstart plans public cloud in low Earth orbit

Last week we talked to a space startup that wants to load satellites with server virtualization tool XenServer so they can run virtual machines (VMs), which sounded intriguing enough that we decided to learn more. And we’re glad we did, because this is a more-than-interesting development in both computing and commercial use of …

  1. Anonymous Coward
    Anonymous Coward

    Call me skeptical but...

    I don't see this taking off.

    Have you ever tried to manually position a dish to pick up a satellite signal in the most optimal way? That can be quite tricky; not everyone has a dish which can automatically align itself. So how is this going to work for a satellite which will only be available for a limited amount of time? Half a centimeter out of alignment can already be a cause for a broken up signal.

    I also can't help wonder that this might be pretty tricky to actually set up for a customer. Once again: if you only got an X amount of time to set things up... Depending on the server and the services you wish to set up you can sometimes be looking at hours of work. Yet it sounds to me as if this system does not provide for that (at least that's the impression I get). After all: those satellites are constantly on the move.

    So what if you're busy setting it up and all of a sudden it moves out of range? Wait for an x amount of hours before you can continue setting things up again?

    1. Pen-y-gors

      Re: Call me skeptical but...

      I suspect they've thought of that...after all, how do the US military talk to their spy satellites?

    2. Pascal Monett Silver badge

      Re: Call me skeptical but...

      I'm just wondering what's the use of a VM that is only active for the minute or two it is circling above you. I don't think there's any communication between satellites, so it's all sat-to-ground and back. I can imagine they'll have several ground stations (and I doubt individuals will be setting up dishes to communicate directly - that does not sound like a good commercial pitch) and they'll be able to manage it all and monitor the fleet continuously, but I still can't find a use case for computing power that regularly disappears.

      Anyone care to enlighten me on that ?

      1. Brangdon

        Re: I don't think there's any communication between satellites,

        There may be communication between satellites. They can aim lasers at each other, for a point-to-point mesh. Again, this is what SpaceX's Internet constellation will do.

        The computing power won't disappear. There will always be a satellite or two overhead, that has the most recent state of your app. That state will be broadcast from satellite to satellite as they move, and your ground station will track the satellites and hand-off from one to the next with solid-state phased arrays. I'm not sure what the applications are, but it does sound cool and workable.

        1. Virtshaun

          Re: I don't think there's any communication between satellites,

          This is exactly how it will work. You can see one of several patents we already have issued on the topic here; with about 20 more in the works. Stay tuned... -shaun coleman

    3. Brangdon

      Re: Call me skeptical but...

      I'm guessing their ground receiver is some kind of phased array. That is, a fixed-position, solid state pizza-box sized array of tiny transceivers. It aims itself by adjusting the timing of how it broadcasts or integrates the signals across the transceivers, and using wave mechanics. Magic, in other words.

      It's what SpaceX plan to use to talk to their global Internet constellation.

    4. DropBear

      Re: Call me skeptical but...

      "Have you ever tried to manually position a dish to pick up a satellite signal in the most optimal way?"

      Have you ever seen a satellite phone? Yes, the data rate of those is kinda crappy, but on the other hand it's only just possible to tell them apart from one of the more modern GSM bricks of yore - they have a slightly larger nub. That's it. Low earth orbit is roughly a hundred times closer than the geostationary stuff we tend to aim dishes at, and things in it only stay on your sky for a few minutes at a time. You don't _aim_ anything at stuff like that unless you're NORAD or something...

    5. Mark 85 Silver badge

      Re: Call me skeptical but...

      I'd almost bet that the company will do the ground comms part also and charge you for the time.

  2. short

    VC Bait?

    This sounds much more like a VC-harvesting pitch than a sane plan. Space is cool, Cloud is cool, and 5 minutes of powerpoint later...

    Good luck to them, I guess.

    1. John Smith 19 Gold badge

      "This sounds much more like a VC-harvesting pitch than a sane plan. "

      What he said.

    2. Christian Berger


      Particularly since investors now have lots of "free" money, as many key interrest rates are now at zero, or close to it.

      So it's only sensible to invest in more speculative things. Of course since VCs don't seem to have a decent technical staff, they cannot easily understand why such a project is ultimately pointless, even if they manage to send some servers into space, as the launch itself costs more than a very decent ground based server.

  3. IanRS

    Blue-sky thinking?

    By the time you are in orbit, the sky is black, not blue.

  4. jake Silver badge

    So, the added complexity & security headaches of "Cloud" ...

    ... but IN SPAAAAAAAACEEE!!!!11!!!!11!!eleven!!!11one!

    Nothing to go wrong there, right?

    Garages in Palo Alto, Nyack, Harrogate, Athens, Wellington, Perth, Cape Town, Rio de Janero and Hilo filled with "breadbox sized" servers will be faster (radiation hardened CPUs are slow), less complex (fewer links in the chain), more secure (see: links), just as redundant, have lower latency, be easier to setup (duh), and be easier to maintain(!) than this hair-brained idea.

    Bottom line: It'll never fly.

    1. Anonymous Coward
      Anonymous Coward

      Ah, the mechanics of orbits

      It'll never fly?

      Technically speaking, if it is in orbit it will be forever falling...

  5. Alister

    That’s some nicely blue-sky thinking, but not the work of space cadets.


  6. NiteDragon

    On top of the rather valid points of the posts above... More space junk and rare minerals burning up in cycles doesn't really sound like the sanest plan ever?

    1. John Smith 19 Gold badge

      "More space junk and rare minerals burning up..doesn't really sound like the sanest plan ever?"

      Like shipping laundry from the Klondike to Hawaii to be washed.

      Yet that was done and people made money doing it.

      You seem to think that sanity is a prerequisite for getting funding or making a profit.

      1. Francis Boyle Silver badge

        Re: "More space junk and rare minerals burning up..doesn't really sound like the sanest plan ever?"

        "Like shipping laundry from the Klondike to Hawaii to be washed."

        Citation please. The only think I could find was an "amusing facts" page which says it was California and China which is a bit to much like a Radio Yerevan to inspire confidence.

        1. jake Silver badge

          Re: "More space junk and rare minerals burning up..doesn't really sound like the sanest plan ever?"

          Re: airing dirty laundry ...

          a.f.u did this one to death about 25 years ago, and again about 15 years ago, along with it's cousin "San Francisco to Canton". It never happened. Logistics say no, and there are no contemporary reports of it happening.

    2. Anonymous Coward
      Anonymous Coward

      As per the article these are only in Low Earth Orbit, so will de-orbit and burn up within a matter of years. Unless someone decides to tow one into a higher orbit, there shouldn't be any chance of long-term debris hanging around.

  7. frank ly

    A computing Network ....

    .... in the sky. In the sky, there will be a network. What shall we call it?

  8. Destroy All Monsters Silver badge

    One runs a real-time OS on ARM or PowerPC silicon and gets the job of keeping the satellite humming. The other will be air-gapped [vacuum-gapped? – Ed] from the main board and will offer “x86 Intel-based, multi-core hardware including several terabytes of high-speed solid state storage.”

    If this circuitry hardened?

    If not, it will be dead soon.

    1. short

      maybe shielding?

      I was wondering how susceptible TBytes of NAND SSD would be, given how much effort it is to make them remember stuff at ground level. This interesting bunch of slides form ESA

      if I followed it right from slide to slide, suggests that, in LEO, and with a 5mm aluminium shield, NAND should be good to go. Maybe LEO, shielding, a short life and a swarm of the damn things for redundancy means that failing hardware won't be the limiting factor here. Sanity, on the other hand...

  9. Anonymous Coward
    Anonymous Coward

    Time for regulation?

    Do we need to start requiring that a workable de-orbiting plan is put in place before things like this are allowed up? I can imagine there will be a lot of junk left behind quite rapidly (failed nodes, companies that go bust).

    1. John Smith 19 Gold badge

      Re: Time for regulation?

      Such requirements already exist and anyone launching from a specific country has to follow that countries plan. Since a lot of payloads have been launched from the US detailed requirements already exist.

      1. Anonymous Coward
        Anonymous Coward

        Re: Time for regulation?

        anyone launching from a specific country has to follow that countries plan

        US plan: "If you are launching from the US, then your satellite must crash-land as far away from the US as possible"

        1. Pedigree-Pete

          ...crash as far away from the US as possible...

          You mean Australia, already been done. PP

  10. Anonymous Coward
    Anonymous Coward

    A couple of problems with this plan

    The main reasons for running software in a satellite are remote sensing and communication relay. Otherwise, the application may as well run in a data centre.

    1a. In the case of sensing: if these are generic satellites, what sort of sensors/cameras do you launch them with? Different users will have very specific sensing needs. As for in-situ VMs, why not just downlink the raw sensing data and process it on the ground? You can then resell the same raw sensing data to multiple customers concurrently and let them all process it in different ways.

    1b. In the case of communication: low-earth constellations have been tried before but have very limited market (e.g. O3B), due to the complexities on the ground side of tracking two satellites across the sky.

    For comms relay, the benefits of running VMs are unclear. I expect you would create a mesh network between the satellites to relay traffic, but that's a low-level service which spans the whole constellation, not something the users will write themselves to run on a single node. What other processing needs to be done in the satellite, which couldn't have been done on the ground?

    2. Even if there is a good commercial reason for microsatellites with mobile VMs, why does the company need to build its own launch vehicle from scratch - a hugely expensive and risky business?

    Is there nothing which already exists that could get a small load into low-earth orbit? And if these satellites are so small, could they not piggy-back on other launches?

    1. John Brown (no body) Silver badge
      Black Helicopters

      Re: A couple of problems with this plan

      "The main reasons for running software in a satellite are remote sensing and communication relay. Otherwise, the application may as well run in a data centre."

      Or maybe extra-legal jurisdiction for certain data processing or tax avoidance?

  11. Denarius Silver badge


    what problem is this hideously expensive scheme solving ? Or have I missed something again?

    1. Anonymous Coward
      Anonymous Coward

      Re: question

      If they have discovered a way to make a reality-bending evil AI running on Intel hardware, it would make sense to keep it hovering over $COUNTRY as literally the Eye of Sauron.

      "Yes. Igor. Throw the switch. The ORBITAL switch. MWAHAHA!"

  12. Mage

    this stuff sounds cool?


    It's nuts as a real world use compared to a boring warehouse full of "pizza box" style computers and storage.

    Exactly what problem does this extremely expensive per MIPS/GByte solution solve?

  13. Mage


    For some reason when I re-read this I thought of DeLorean.

    It got lots of money. He told the UK the Irish Government / Shannon Development was going to approve it for the old Firenca plant at Limerick (They actually thought it was doomed or a scam and chased him).

    The cars were flimsy, road vibrations caused the thin stainless steel skin to de-laminate from the real fibregrass body (thus not scaleable production). The engine was more suitable power for a regular car than a high end sports car. The cool gull wing doors were a hazard if it flipped.

    It looked cool, but was the sort of thing better suited to a film set than real life.

    Was he a scammer or simply stupid or something else? I don't know. He offered to site in West Belfast and create LOADS of jobs, so the UK continued to starve indigenous startups and dished him loads of money.

    If this gets of the ground ( ^_^ ) it will fail.

    Satellites don't need the overhead of a VM.

    It's an irresponsible use of LEO.

    1. Francis Boyle Silver badge

      Please teacher, I know this one.

      He was a scammer.

      Next question.

    2. jake Silver badge

      Re: VC

      You forgot "It was rear-engined and handled worse than a 911 at speed" (hard as that is to believe!). For small values of "speed", of course.

  14. M7S

    All that hardware in orbit, bare metal or not

    would that be like an iron sky?

    It's Friday, but my choice of DVD tonight has now been....dictated.

  15. John Smith 19 Gold badge

    So you're counter rotating the apps against the sat movement to keep a pseudo server overhead.

    Which is damm tricky. Major kudos to any team that can make it work. Hardware wise the phrase South Atlantic Anomaly looms large. SpaceX took a while to devise radiation tolerant (IE resistant by architecture), rather than radiation hard (mfg in rad hard, v. expensive process) computer systems. And will those Intel based boards carry the epicly easy to access AMT processor (I'd call it a hack, but just-leave-login-details-blank doesn't really qualify) to manage this?

    BTW a note on times. Time running on any particular server will depend on how long the bird is visible from the ground station. A full orbit at this sort of altitude is about 90mins so you could have 10s of mins if it's one bird/customer (which sounds unlikely). But how will you transfer context of running job from satellite to satellite?

    3 LEO/MEO comms constellations were launched in the 90's. All are still running but all went through Chp11. Orbcomm does low(ish) bandwidth M2M and Iridium survives on the block buy of the USG.

    I like optimism, but there's optimism and there's Forest Gump optimism. So let's see how much VC they can get.

  16. Allan George Dyer

    "The Cloud Crashed"

    Will have a whole new meaning.

  17. dbayly

    Possible motivation

    Could it be that the servers are out of reach of any regulator or other authority , that makes this an attractive proposition for some?

    1. annodomini2

      Re: Possible motivation

      Also being on orbit, harder to enforce, e.g. bypass the Great firewall as it could be satellite accessed.

  18. Anonymous Coward
    Anonymous Coward

    "out of reach of any regulator or other authority"

    I was going to suggest that, but it didn't stop the US demanding that Microsoft provide access to information stored in Ireland - it will still come back to who owns the birds.

  19. herman Silver badge

    Theirs is a dumb idea. Better to keep the satellites simple bent pipes and keeping all the VMs on the ground stations. Then you don't need to migrate anything, just switch carriers. I guess that is too simple for them excite vulture capital.

  20. actual space mission engineer

    A few clarifying comments to the above thread:

    The key reason for space based virtualization is timeliness - right now storing and dumping raw data results in yesterday's data tomorrow. Using the VM allows for the ability to move some of the ground processing software up to the vehicle, to either preprocess radar data or at the very least determining which images are just of clouds and not worth wasting the bandwidth to downlink.

    Radiation is an issue, but you need to think of the VM as a "mission coprocessor" like an old math coprocessor. The VM is not running the satellite (spinning wheels, firing thrusters, orienting the vehicle, health & status, GPS, etc.), it is just (pre)processing payload data. The VM hardware will be shielded (either aluminum as stated above or high z materials) and as with virtualization can be restarted from a previous checkpointed/saved state after a fault.

    As the mission progresses new code can be loaded to the VM to further exploit the data, without worrying about interfering with the flight control software or hardware.

    Extending this concept across the constellation would allow for enabling "virtual persistence" of a target region, with each satellite handing off its current VM state or subset of data to the next satellite about to rise over the area. Also allows for satellites to become multi-use vehicles, either over an orbit or throughout their lifetime.

    With respect to overflight time, 10-15 minutes is what a LEO vehicle has to a groundsite - been that way since Sputnik - its just physics. All contacts between the ground and the vehicle are worked out like a TV schedule, with the vehicle turning on a few seconds before the overflight so the ground can lock up on the signal. The real issue is the frequency - at UHF you can be 30 degrees wrong and still find the signal, at S-band its about 5 deg and Ka 0.1 deg. TLEs were made to point S-band antennas - higher frequencies like X and Ka require the owner operator to perform their own precision orbit determination to guarantee the link closes and every second of a pass is used.

  21. allthecoolshortnamesweretaken

    Okay, after re-reading the article and reading the thread I think I understand the how.

    What I still do not understand is the why.*

    This reminds me of the Iridium satellite phone failure. $5 Bn spent to build and launch the infrastructure. Projections made in 1998 said: 500,000 subscribers by 1999, profit. Reality in 1999: 10,000 subscribers, Iridium files for bankruptcy. Which, looking back, feels a bit like the starting signal for The Big DotCom CalamityTM.

    * Other than 'it's cool to build your own rockets and satellites' which, let's admit it, it is.

    1. Chairman of the Bored

      I like the Iridium comparison as in many ways Iridium can be thought of as this 'seamlessly shared LEO resource' similar to that proposed. The Iridium system architecture is beautiful, and Motorola (may she rest in peace) did a fantastic job applying mass production techniques to bring the space segment cost under control.

      The problem I think is that Iridium hit the street at the same time that the GSM Juggernaut really took off. You're still talking about many billions dropped in infrastructure, but you don't have to pay it all up front, the handsets are cheaper, you can do partial steps to useful capability, pain is spread between many stakeholders, etc. The market voted and its pretty obvious who one

      I've used Iridium in some truly bizarre off-grid places ... Where the cell towers absolutely will never grow... And it gets the job done. Much less of a pain in the ass than Inmarsat terminals. But how often is that a relevant use case?

      What I find much more interesting is the Globalstar architecture. Rather than cover the whole planet through crosslinked SV, Globalstar's is a bent pipe solution using a GSM air interface. The philosophy here is 'why cover the whole planet when the vast majority of the population lies within a SV footprint width of a GSM tower... Makes for very, very simple and inexpensive SV. You lose coverage in polar regions, middle of ocean, but these are special cases

      BUT - there is such a thing as being too cheap. Globalstar skimped on testing (or something!) The s-band downlink subsystems on the SVs couldn't hack the space weather environment and wet titsup. As the L- bank uplink still worked, Globalstar became a one way asset tracking system - staple of long haul hazmat trucking

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like