Author Topic: Amazon Web Services hiring engineers for “big, audacious space project”  (Read 2875 times)

Offline Johnnyhinbos

  • Full Member
  • ****
  • Posts: 1373
  • Boston, MA
  • Liked: 1593
  • Likes Given: 222
John Hanzl. Author, action / adventure www.johnhanzl.com

Offline dchill

  • Member
  • Posts: 90
  • Liked: 12
  • Likes Given: 0
CNBC had posted an article last night: https://www.cnbc.com/2018/09/11/amazons-cloud-business-is-looking-to-space.html
AWS is already being used by other parts of the US Gov.

Offline ChrisWilson68

  • Senior Member
  • *****
  • Posts: 3700
  • Sunnyvale, CA
  • Liked: 2277
  • Likes Given: 2917
Everyone already knows that Amazon is testing drone aircraft to deliver packages to people in an hour.

This "big, audacious space project" must be the next logical step: delivery from space.  Having to wait an hour is far too long.  With thousands of LEO satellites in orbit, Amazon will be able to deliver your package to you in minutes.

Strangely, defense officials are also interested in this technology, though it's unclear why.

Online TrevorMonty

Bezos has spoken often about developing space infrastucture to allow low cost access to space. Up till now its low cost launch but looks like he is also tackling communications in form of space internet
He goal is to enable anybody with right skills and drive to start space equivalent of internet company (Amazon, facebook, ebay, google etc) from their garage without needing $10-1000s millions.

For example with space internet earth observation satellites could off load their photo data to another satellite within seconds of being taken, instead of to waiting for next free ground station. Would be no limits on number of photos taken per orbit. With in space cloud services only processed images may end up being sent back to earth.

Expect another communication satellite constellation but aimmed at servicing satellites and space vehicles, not broadband to remote parts of ground. Unlike other constellations, funding shouldn't  be an issue.
« Last Edit: 09/12/2018 08:54 PM by TrevorMonty »

Offline Bynaus

  • Full Member
  • ****
  • Posts: 496
  • Planetary Scientist
  • Switzerland
  • Liked: 343
  • Likes Given: 238
Server farms in space? Long-term stable and secure environment, and you radiate against the microwave background (aka the best heat sink in the universe). Okay, you have cosmic rays, but there are ways to check against that (see SpaceX's solutions for Dragon, which uses off-the-shelf electronics). And MOD damage, but even that can be mitigated somewhat by being far above LEO.

Offline butters

  • Full Member
  • ****
  • Posts: 1762
  • Liked: 387
  • Likes Given: 134
Sounds like ground-segment-as-a-service. Customers own the satellites but AWS owns the ground stations and provides abstracted services for fleet management, downlink/uplink, and common data processing tasks (e.g. splicing overlapping image data from earth observation satellites).

I don't think this is about deploying cloud computing resources in space or competing with Starlink/OneWeb as a space-based network operator, although the latter is less unlikely than the former.

Offline Asteroza

  • Full Member
  • ****
  • Posts: 626
  • Liked: 81
  • Likes Given: 2
Oh, that's gonna scare the pants off all those timeshare ground network services startups. Why schedule slots on a startup when you can schedule timeslots on AWS and direct pipe downlinks that feed directly to AWS cloud services on favorable network routes.

One can also see this feeding into the Lambda Edge cloud services, deploying single container datacenters into suburban areas underserved by incumbent network providers by bypassing them for longhaul, as the incumbents make more money at their network edges where they peer for transit.

Evil endgame thought is that most current cellular wireless providers don't have the capital tables to provide proper 5G deployments (Francis McKierney talks about this in his concept of future infrastructure being functionally an edgeless all enveloping cloud system), but if Amazon was starting from scratch, they can do the combo of containerized datacenters (containing CDN services and edge computing) combined with 5G RF/baseband with backhaul by via LEO megaconstellation satellite to completely short circuit the current cloud to end user paradigm. Not clear if they could get away with moving the 5G baseband to the datacenter completely though (as I understand it, datacenter centralized baseband uses fiber optics to maintain tight timing control, allowing the entire cellular network to act like an intelligent phased array antenna system and keeping cell tower equipment costs down by limiting to antennas and RF components only without a full baseband processor, though chip scale atomic clocks might make it workable without fiber optics)

Offline DigitalMan

  • Full Member
  • ****
  • Posts: 532
  • Liked: 91
  • Likes Given: 9
Oh, that's gonna scare the pants off all those timeshare ground network services startups. Why schedule slots on a startup when you can schedule timeslots on AWS and direct pipe downlinks that feed directly to AWS cloud services on favorable network routes.

One can also see this feeding into the Lambda Edge cloud services, deploying single container datacenters into suburban areas underserved by incumbent network providers by bypassing them for longhaul, as the incumbents make more money at their network edges where they peer for transit.

Evil endgame thought is that most current cellular wireless providers don't have the capital tables to provide proper 5G deployments (Francis McKierney talks about this in his concept of future infrastructure being functionally an edgeless all enveloping cloud system), but if Amazon was starting from scratch, they can do the combo of containerized datacenters (containing CDN services and edge computing) combined with 5G RF/baseband with backhaul by via LEO megaconstellation satellite to completely short circuit the current cloud to end user paradigm. Not clear if they could get away with moving the 5G baseband to the datacenter completely though (as I understand it, datacenter centralized baseband uses fiber optics to maintain tight timing control, allowing the entire cellular network to act like an intelligent phased array antenna system and keeping cell tower equipment costs down by limiting to antennas and RF components only without a full baseband processor, though chip scale atomic clocks might make it workable without fiber optics)

Didn't the article mention maintaining an active top secret clearance?  The 'project' is likely the $10 billion DOD contract.

Offline Lar

  • Fan boy at large
  • Global Moderator
  • Senior Member
  • *****
  • Posts: 10451
  • Saw Gemini live on TV
  • A large LEGO storage facility ... in Michigan
  • Liked: 7315
  • Likes Given: 5104
Server farms in space? Long-term stable and secure environment, and you radiate against the microwave background (aka the best heat sink in the universe). Okay, you have cosmic rays, but there are ways to check against that (see SpaceX's solutions for Dragon, which uses off-the-shelf electronics). And MOD damage, but even that can be mitigated somewhat by being far above LEO.

There was a whole thread on server farms in space (in LEO) a while back, didn't find it in a really quick search but it's out there. Conclusion at the time was that serving content might not make a lot of sense since what content is popular varies geographically meaning as the birds move, they have the wrong content, and that computations might not be a

My guess is that AWS is the logical division (it's probably not a good fit for movie development, for example, LOL) but that the likely product is more of a Starlink competitor. But it might well have server aspects.

Delivery? Funny idea but not yet. 20-30 years out it might be.

The top secret clearance means that some of the traffic might be DOD related.

All IMHO of course.
« Last Edit: 09/13/2018 05:49 PM by Lar »
"I think it would be great to be born on Earth and to die on Mars. Just hopefully not at the point of impact." -Elon Musk
"We're a little bit like the dog who caught the bus" - Musk after CRS-8 S1 successfully landed on ASDS OCISLY

Offline Asteroza

  • Full Member
  • ****
  • Posts: 626
  • Liked: 81
  • Likes Given: 2
If the security clearance is not a red herring, then I could make similar arguments with the ground terminal container dovetailing with DARPA's project blackjack and AWS GovCloud via FOB forward deployment, plus mix&match with tactical BACN pod deployments. Security isn't much of an issue in this day and age with IPSEC VPN deployments and things like Unisys' Stealth VPN network software.

Online sanman

  • Senior Member
  • *****
  • Posts: 4335
  • Liked: 649
  • Likes Given: 8
Cloud in Space?

Or would it mainly be Mesh Network in Space, with Cloud communicating across that Mesh Network?

Offline Chasm

  • Full Member
  • ***
  • Posts: 370
  • Liked: 160
  • Likes Given: 0
I'd say Mesh Network first. In the cloud world you have to communicate, a lot. Companies usually find out really quick that they have to upgrade their internet connection once they moved into the cloud. There is also a lot more compute on planet. I'd try really hard to communicate both on orbit and with all the US Bases, not just the big sat com centers.

Next up? ASAP stick an AWS Snowball Edge into a sat or two. Local data collection and some AWS compute functions. The military is apparently already using them so nothing new but the location and the speed they move at. :D  "Just" convert them for in space operation... The follow on generations can be more specialized.


I would say that AWS is space is for the most part a LEO thing. Any relevant amount of storage or compute required commercial technology, rad hardened gear is just way too slow. Stay low for protection from the magnetic field, don't expect a decade long live span low and replace often.
Parts of the technology, esp. mesh networks, can be used in other places.

Offline Lar

  • Fan boy at large
  • Global Moderator
  • Senior Member
  • *****
  • Posts: 10451
  • Saw Gemini live on TV
  • A large LEGO storage facility ... in Michigan
  • Liked: 7315
  • Likes Given: 5104
If launch costs are lower, using closer to off the shelf gear and planning to replace it much faster than conventionial sats is completely congruent with how AWS, Google, IBM, et al equip their terrestrial data centers.
"I think it would be great to be born on Earth and to die on Mars. Just hopefully not at the point of impact." -Elon Musk
"We're a little bit like the dog who caught the bus" - Musk after CRS-8 S1 successfully landed on ASDS OCISLY

Offline Asteroza

  • Full Member
  • ****
  • Posts: 626
  • Liked: 81
  • Likes Given: 2
I'd say Mesh Network first. In the cloud world you have to communicate, a lot. Companies usually find out really quick that they have to upgrade their internet connection once they moved into the cloud. There is also a lot more compute on planet. I'd try really hard to communicate both on orbit and with all the US Bases, not just the big sat com centers.

Next up? ASAP stick an AWS Snowball Edge into a sat or two. Local data collection and some AWS compute functions. The military is apparently already using them so nothing new but the location and the speed they move at. :D  "Just" convert them for in space operation... The follow on generations can be more specialized.


I would say that AWS is space is for the most part a LEO thing. Any relevant amount of storage or compute required commercial technology, rad hardened gear is just way too slow. Stay low for protection from the magnetic field, don't expect a decade long live span low and replace often.
Parts of the technology, esp. mesh networks, can be used in other places.

The problem point here is while you might have mesh access (200+ sats in line of sight for the megaconstellations), you are going to have to do some pretty impressive phased array transmission at the ground terminal to fire RF packets at every sat simultaneously plus the added headaches of multipath TCP (though iPhones can do that now) plus the headaches of active dynamic mesh routing (see the teeth gnashing over BATMAN and ROBIN protocols) to get enough bandwidth for Serious Big Data piping to cloud resources elsewhere on the planet. You start getting at the weird balance between sorta bursty big-ish data with a lowercase b, and ground to orbit lasercomm for Big Data with an uppercase B.

Offline as58

  • Full Member
  • ****
  • Posts: 802
  • Liked: 276
  • Likes Given: 179
Server farms in space makes no sense to me. A server farm may use tens of MW and getting power and especially cooling would be much harder than on Earth.

Online meekGee

  • Senior Member
  • *****
  • Posts: 8495
  • N. California
  • Liked: 4685
  • Likes Given: 884
Server farms in space makes no sense to me. A server farm may use tens of MW and getting power and especially cooling would be much harder than on Earth.

And server farms try to be more local to where the customers are.   An orbital server farm spends most of its time over the pacific or the poles and needs a lot of extra communication bandwidth to make up for it.

Plus as you say power and cooling, plus the cost of maintenance and upgrades...

I just don't see it either.

I'm betting on an alternate constellation.  OneWeb (if it ever gets off the ground) can't compete with StarLink, and NG needs customers.

To me, with how things are, I can't see how Bezos hasn't gone after a constellation already.
ABCD - Always Be Counting Down

Offline matthewkantar

  • Full Member
  • ****
  • Posts: 829
  • Liked: 535
  • Likes Given: 640
Unlimited power and free real estate though.

Offline TripleSeven

  • Full Member
  • ****
  • Posts: 639
  • Istanbul turkey
  • Liked: 268
  • Likes Given: 1108



To me, with how things are, I can't see how Bezos hasn't gone after a constellation already.

they are a pretty big gamble.  Iridium finally is paying off, but it took some significant technological breakthroughs and essentially wiping away the debt from the original constellation to make money

The Gulf war and the various mideast wars have "helped" because there is a large military component to Iridium...and  they are starting to branch out with Airline services (including ADSB) ...but mostly they are still "institutions" not people.

essentially what a constellation of small sats banks on is that people like me, can get better internet from "them" then we can, at fixed locations get from local providers.  any constellation is going to have to have a fairly sophisticated tracking antenna, which means rooftop installations in worlds particularly in the US where outside antennas are frowned on.

the notion of "more people plugging in" is great...except the people who are not plugged in "now" at high speed are mostly not plugged in for a reason...cost.  now part of that is cost of infrastructure...but having spent time in  Bangladesh,  India and Africa (Nigeria/South Africa) the number of people who have the funds to pay much above the 30 USD per month cost is fairly small (ie most people there get internet on their phones at very slow rates)  I dont see the Chinese being part of any of this...

if the product can beat say 100 mpbs at around 50 dollars a month...it would sale.  right now in Houston I can get 1000 MPBS for 140 a month. we have [email protected] fiber and its growing...and we are kind of rural.

here in Istanbul I get 475 MPBS for about 60 dollars a month and fiber is about three blocks over and coming in either October or November.

I am not telling you it wont work...but the numbers are going to have to be competitive.

 
« Last Edit: 09/23/2018 08:32 PM by TripleSeven »

Tags: