Author Topic: Amazon Project Kuiper Broadband Constellation  (Read 195007 times)

Offline ncb1397

  • Senior Member
  • *****
  • Posts: 3497
  • Liked: 2310
  • Likes Given: 29
Re: Amazon Project Kuiper Broadband Constellation
« Reply #120 on: 10/03/2019 01:15 am »

There's been discussions of putting some sort of cache for content in the constellation satellites themselves to move things like movies and other streaming content closer to the users.  I could see how something like having a bunch of SSDs to cache content in satellites could be useful.  But not much beyond that.

You basically just described a data center.

I work with our own data centers across the globe daily, so this is actually in my wheelhouse.  What I've described is more like a stripped down CDN and not a data center.  Data centers provide compute and storage among other services.  If Starlink provides basic CDN features to cache content closer to customer downlinks, it would be a huge advantage and prevent a lot of duplicate data from having to traverse their constellation as well as very extremely low latency.  Whether they'll get into that game or not and solve all the technical issues with delivering content directly from their birds and all the associated hand-off issues is unknown.  I'd put the chance of that as pretty unlikely for the first few iterations of Starlink as it really is an optimization they can tackle after they've got the basics and ISLs working.

Anyways, data repository in space has already been done. Do you think that when you order video on demand from Dish or DirectTV, they upload the file to the satellite, let you watch it, and then delete it only for the next user to request it shortly thereafter? No, that would be silly.
When you say "has already been done", is this based on anything you actually know?

Because what you're describing doesn't sound to me like something that actually exists or is practical.

Well, if you do the math, you can fit about 40,000 2 hour long HD videos on a single 100 TB SSD that likely costs well under $100,000. That should be comparable to Netflix's catalog which uses up 15% of internet bandwidth(and will presumably use up a similar amount on a NGSO constellation). As far as compute costs, the top end GPU from nVidia costs $2500 and weighs ~1.3 kg. Assuming $5000 per kg, that puts the cost delivered on orbit at $9000 - about 3-4x the cost. As far as utilities, assuming $5000 per kg and 100 W per kg, cost for utilities in LEO would be about $1 per kwh over a 10 year service life - again about 3x the cost of electricity in Germany.

But the major issue with satellite internet has been bandwidth caps that essentially is destroyed by video. This could be one strategy to get around that by at least minimizing the impact on the up link side.

Online meekGee

  • Senior Member
  • *****
  • Posts: 14667
  • N. California
  • Liked: 14670
  • Likes Given: 1420
Re: Amazon Project Kuiper Broadband Constellation
« Reply #121 on: 10/03/2019 02:50 am »

Anyways, data repository in space has already been done. Do you think that when you order video on demand from Dish or DirectTV, they upload the file to the satellite, let you watch it, and then delete it only for the next user to request it shortly thereafter? No, that would be silly.
When you say "has already been done", is this based on anything you actually know?

Because what you're describing doesn't sound to me like something that actually exists or is practical.

Well, if you do the math, you can fit about 40,000 2 hour long HD videos on a single 100 TB SSD that likely costs well under $100,000. That should be comparable to Netflix's catalog which uses up 15% of internet bandwidth(and will presumably use up a similar amount on a NGSO constellation). As far as compute costs, the top end GPU from nVidia costs $2500 and weighs ~1.3 kg. Assuming $5000 per kg, that puts the cost delivered on orbit at $9000 - about 3-4x the cost. As far as utilities, assuming $5000 per kg and 100 W per kg, cost for utilities in LEO would be about $1 per kwh over a 10 year service life - again about 3x the cost of electricity in Germany.

But the major issue with satellite internet has been bandwidth caps that essentially is destroyed by video. This could be one strategy to get around that by at least minimizing the impact on the up link side.

So basically, you're assuming it has been done since your calculations argue that it will make sense.

But reality is so much more complicated than multiplying a bunch of spec sheet parameters... 

so when you say "XXX has been done", you're creating a false impression.  You don't know that it has, you're WAGging that it has.

It could very well be that since this solution will at most save 50% of bandwidth (since down-streams still have to occur in the same way) and since up streams can utilize a higher bandwidth narrow beam, there isn't enough value in this proposition for anyone to bother.  Or not.  I don't know. 

I know that when I google "content caching on satellites" I get a long list of papers, not a long list of satellite companies.  Maybe there's a hint there.
ABCD - Always Be Counting Down

Offline octavo

  • Full Member
  • **
  • Posts: 247
  • Liked: 186
  • Likes Given: 740
Re: Amazon Project Kuiper Broadband Constellation
« Reply #122 on: 10/08/2019 09:04 am »
When you say "has already been done", is this based on anything you actually know?

Because what you're describing doesn't sound to me like something that actually exists or is practical.

The major problem with this idea is lack of serviceability. SSD's have limited lifespan, and with thousands of reads per second 24/7 it wont take long for them to start failing. Even raided, once enough disks fail, the entire array will die.
Traditional HDD's have a longer life, but are slower and I assume that thousands of metal platters spinning at various speeds may cause some control/pointing issues.

I don't see this working out well. You might also need rad-shielding. NAND Flash is pretty volatile.


Offline launchwatcher

  • Full Member
  • ****
  • Posts: 765
  • Liked: 729
  • Likes Given: 996
Re: Amazon Project Kuiper Broadband Constellation
« Reply #123 on: 10/09/2019 07:12 pm »
When you say "has already been done", is this based on anything you actually know?

Because what you're describing doesn't sound to me like something that actually exists or is practical.

The major problem with this idea is lack of serviceability. SSD's have limited lifespan, and with thousands of reads per second 24/7 it wont take long for them to start failing.
The usual problem with SSD's is write endurance rather than read endurance; unlike DRAM, flash memory doesn't require refresh on read.

That said, I agree that LEO caching for CDNs doesn't make sense with current mass/power/cooling budgets, in part because you lose any sense of geographic locality in LEO.   

Much better to relay to ground-based caches a hop or two from the customers, which would benefit from geographic locality, would not be mass constrained and in underdeveloped areas could conceivably be built to be entirely standalone, requiring only electrical power and perhaps chilled water inputs.   

Look at what you could pack into a single 20' or 40' shipping container and deliver anywhere on the planet that needs the cache..

Offline thirtyone

  • Full Member
  • **
  • Posts: 256
  • Liked: 431
  • Likes Given: 354
Re: Amazon Project Kuiper Broadband Constellation
« Reply #124 on: 10/09/2019 08:44 pm »
I've worked with low level NAND devices before (without controllers, ECC, etc.) My first impression is that any recent NAND memory can't be sent to space, at all, ever.

I think most people would be appalled by how miserably bad those arrays are. They actually do very occasionally lose bits during reads as the charge storing the bits leaks out - you don't hear about these things until you find the industry presentations. The raw bit error rate of some of the modern 3D NAND is easily 1 in 1000 (I'm not kidding here). People are literally continually developing new ECC algorithms to deal with the increasing error rates of higher and higher density NAND.

Actually quite impressive they can convert a device with such awful reliability into something people consider to be better than HDDs. Kind of scary if I think about it too hard, though. There's probably quite a bit of planned obsolescence the businesses in this market are planning on...

Anyway, to the point - imagine such a high error rate memory getting sent into space, where it'd have to deal with all the extra radiation. I'm...pretty sure any modern density NAND is basically completely unusable, even with absurd amounts of ECC. I think they'd have to step a few generations back, and then apply modern ECC to that drive. Here's a space vendor who sells a built NAND system:

https://www.mrcy.com/siteassets/product-datasheets/ssd/trrust-stor-vpx-rt-rad-tolerant-datasheet.pdf

This is a bit larger than a 3.5" HDD, and holds 440GB. Manufacturer claims "<500g." Don't even want to know the price...

Some more info on NAND in space:
https://passive-components.eu/cots-in-space-products-vs-components-for-space-applications/

Offline Ronsmytheiii

  • Moderator
  • Global Moderator
  • Senior Member
  • *****
  • Posts: 23394
  • Liked: 1880
  • Likes Given: 1045
Re: Amazon Project Kuiper Broadband Constellation
« Reply #125 on: 12/08/2019 09:14 am »

That said, I agree that LEO caching for CDNs doesn't make sense with current mass/power/cooling budgets, in part because you lose any sense of geographic locality in LEO.   

Of course, if one were to have a massive payload fairing volume and cheap reusable lifter you can start to chip away at those issues.....

Offline Ronsmytheiii

  • Moderator
  • Global Moderator
  • Senior Member
  • *****
  • Posts: 23394
  • Liked: 1880
  • Likes Given: 1045
Re: Amazon Project Kuiper Broadband Constellation
« Reply #126 on: 12/19/2019 05:13 am »
https://twitter.com/thesheetztweetz/status/1207394184178651138


Edit: Amazon website with the announcement

https://blog.aboutamazon.com/company-news/amazon-building-project-kuiper-headquarters-in-redmond-wa

Quote
As Amazon’s Project Kuiper team continues to make great strides, the team has once again outgrown its current facility. With that in mind, we are leasing and renovating a long term home for the Kuiper team in Redmond, WA, which will become Kuiper’s primary headquarters for research & development, as well as its primary prototype manufacturing and qualification facility. Amazon Kuiper’s new state-of-the-art facility will consist of two buildings with a total of 219,000 square feet of space. It will include offices and design space, R&D labs, and prototype manufacturing facilities. Renovations on the facility are already underway, and the Kuiper team will move into the new site in 2020
« Last Edit: 12/19/2019 05:16 am by Ronsmytheiii »

Offline Rondaz

  • Senior Member
  • *****
  • Posts: 27059
  • Liked: 5301
  • Likes Given: 169
Re: Amazon Project Kuiper Broadband Constellation
« Reply #127 on: 12/20/2019 03:36 am »
SpaceX Is Lobbying Against Amazon’s Internet-Beaming Satellites

Amazon is trying to get a waiver to FCC rules that companies like SpaceX and OneWeb had to follow.

By Todd Feathers
Dec 19 2019, 6:00am

https://www.vice.com/en_us/article/5dmzyx/spacex-is-lobbying-against-amazons-internet-beaming-satellites?utm_source=reddit.com
« Last Edit: 12/20/2019 11:08 am by Chris Bergin »

Offline TrevorMonty

Re: Amazon Project Kuiper Broadband Constellation
« Reply #128 on: 02/06/2020 07:14 pm »
Jeff Foust (@jeff_foust) tweeted at 6:41 AM on Fri, Feb 07, 2020:
Boggett suggests Amazon’s proposed Project Kuiper broadband constellation could be something of a loss-leader: they can fund its development but without a need for a direct return since it will enable other parts of their overall business. #smallsatshow
(https://twitter.com/jeff_foust/status/1225474669932838914?s=09)

This is Mark Boggett of Seraphim Capital.

I think he is right, the network would give Amazon a private and secure internet network for their AWS data centres and customers. Doesn't mean Amazon won't go after consumer internet market but it maybe more of sideline than primary focus of this network.

AWS is profitable enough to fund this constellation on its own.
« Last Edit: 02/06/2020 07:14 pm by TrevorMonty »

Offline FutureSpaceTourist

  • Global Moderator
  • Senior Member
  • *****
  • Posts: 50668
  • UK
    • Plan 28
  • Liked: 85173
  • Likes Given: 38157
Re: Amazon Project Kuiper Broadband Constellation
« Reply #129 on: 04/02/2020 08:50 pm »
https://twitter.com/b0yle/status/1245807745552363521

Quote
In its latest petition against #ProjectKuiper's @FCC filing @SpaceX argues @amazon requests special treatment by inclusion in a processing round closed 3y ago, evading commitments towards interference avoidance and a "mechanical approach" towards the same: https://licensing.fcc.gov/myibfs/download.do?attachment_key=2247176

Offline Nilof

  • Full Member
  • ****
  • Posts: 1173
  • Liked: 593
  • Likes Given: 707
Re: Amazon Project Kuiper Broadband Constellation
« Reply #130 on: 04/24/2020 08:47 pm »
I've worked with low level NAND devices before (without controllers, ECC, etc.) My first impression is that any recent NAND memory can't be sent to space, at all, ever.

I think most people would be appalled by how miserably bad those arrays are. They actually do very occasionally lose bits during reads as the charge storing the bits leaks out - you don't hear about these things until you find the industry presentations. The raw bit error rate of some of the modern 3D NAND is easily 1 in 1000 (I'm not kidding here). People are literally continually developing new ECC algorithms to deal with the increasing error rates of higher and higher density NAND.

Actually quite impressive they can convert a device with such awful reliability into something people consider to be better than HDDs. Kind of scary if I think about it too hard, though. There's probably quite a bit of planned obsolescence the businesses in this market are planning on...

Anyway, to the point - imagine such a high error rate memory getting sent into space, where it'd have to deal with all the extra radiation. I'm...pretty sure any modern density NAND is basically completely unusable, even with absurd amounts of ECC. I think they'd have to step a few generations back, and then apply modern ECC to that drive. Here's a space vendor who sells a built NAND system:

https://www.mrcy.com/siteassets/product-datasheets/ssd/trrust-stor-vpx-rt-rad-tolerant-datasheet.pdf

This is a bit larger than a 3.5" HDD, and holds 440GB. Manufacturer claims "<500g." Don't even want to know the price...

Some more info on NAND in space:
https://passive-components.eu/cots-in-space-products-vs-components-for-space-applications/

NAND is solid state and mechanically reliable though, and error correction solves most of its problems.

Yes, smaller circuits mean higher error rate, but using a higher density NAND array with a layer of ECC that puts you back at the density of the old one still puts you way ahead in terms of reliability.
For a variable Isp spacecraft running at constant power and constant acceleration, the mass ratio is linear in delta-v.   Δv = ve0(MR-1). Or equivalently: Δv = vef PMF. Also, this is energy-optimal for a fixed delta-v and mass ratio.

Offline Semmel

  • Senior Member
  • *****
  • Posts: 2178
  • Germany
  • Liked: 2433
  • Likes Given: 11922
Re: Amazon Project Kuiper Broadband Constellation
« Reply #131 on: 06/30/2020 12:16 pm »
Does anyone has an idea of the rough characteristics of the satellites?

Size, weight, shape?

Also, would be interested if the info from here is still valid, i.e.

Quote
The filings lay out a plan to put 3,236 satellites in low Earth orbit — including 784 satellites at an altitude of 367 miles (590 kilometers); 1,296 satellites at a height of 379 miles (610 kilometers); and 1,156 satellites in 391-mile (630-kilometer) orbits.

and

Quote
Amazon said the satellites would provide data coverage for spots on Earth ranging in latitude from 56 degrees north to 56 degrees south. About 95 percent of the world’s population lives within that wide swath of the planet.

Offline Asteroza

  • Senior Member
  • *****
  • Posts: 2910
  • Liked: 1126
  • Likes Given: 33
Re: Amazon Project Kuiper Broadband Constellation
« Reply #132 on: 07/01/2020 12:15 am »
Amazon formally announces space business solutions business unit...

https://blog.aboutamazon.com/innovation/amazon-web-services-unveils-new-space-business-segment


Your one stop end-to-end shop for your platform needs...


Also, a ton of new Project Kuiper related jobs along with the new business unit (unfortunately the acronym is problematic...) related jobs are up on Amazons job portal now, along with engineers privately calling out for interested people to apply on LinkedIn and elsewhere. Definitely shifting into a faster gear (rather than Blue Origin's turtle pace)...
« Last Edit: 07/03/2020 07:49 am by Asteroza »

Offline Ronsmytheiii

  • Moderator
  • Global Moderator
  • Senior Member
  • *****
  • Posts: 23394
  • Liked: 1880
  • Likes Given: 1045
Re: Amazon Project Kuiper Broadband Constellation
« Reply #133 on: 07/03/2020 08:01 am »
Amazon formally announces space business solutions business unit...

https://blog.aboutamazon.com/innovation/amazon-web-services-unveils-new-space-business-segment


Your one stop end-to-end shop for your platform needs...

So notice how it is under the AWS division, I think that is a strong indicator that Kuiper's biggest client will be Amazon/AWS itself, and will mostly make money off of new AWS subscriptions. This is a pretty big leg up on the other constellations, as it is business/enterprise oriented and not consumer facing primarily and therefore has a must stable initial revenue service. Plus it makes AWS much more competitive for future Government contracts.

Online gongora

  • Global Moderator
  • Senior Member
  • *****
  • Posts: 10435
  • US
  • Liked: 14349
  • Likes Given: 6144
Re: Amazon Project Kuiper Broadband Constellation
« Reply #134 on: 07/03/2020 03:17 pm »
The rocket program isn't Amazon.

Online matthewkantar

  • Senior Member
  • *****
  • Posts: 2189
  • Liked: 2647
  • Likes Given: 2314
Re: Amazon Project Kuiper Broadband Constellation
« Reply #135 on: 07/03/2020 03:28 pm »
And BO started before SpaceX.

Offline Nomadd

  • Senior Member
  • *****
  • Posts: 8894
  • Lower 48
  • Liked: 60677
  • Likes Given: 1333
Re: Amazon Project Kuiper Broadband Constellation
« Reply #136 on: 07/03/2020 04:03 pm »
And BO started before SpaceX.
The company started as a vague idea.
Those who danced were thought to be quite insane by those who couldn't hear the music.

Offline Nomadd

  • Senior Member
  • *****
  • Posts: 8894
  • Lower 48
  • Liked: 60677
  • Likes Given: 1333
Re: Amazon Project Kuiper Broadband Constellation
« Reply #137 on: 07/03/2020 04:04 pm »
The rocket program isn't Amazon.
  I meant the man. Just my brain trying to think and chew gum at the same time.
« Last Edit: 07/03/2020 04:05 pm by Nomadd »
Those who danced were thought to be quite insane by those who couldn't hear the music.

Offline johnlandish

  • Member
  • Posts: 38
  • Europe
  • Liked: 105
  • Likes Given: 5
Re: Amazon Project Kuiper Broadband Constellation
« Reply #138 on: 07/04/2020 11:46 am »
And BO started before SpaceX.

As a tiny research company for the first few years.

Offline su27k

  • Senior Member
  • *****
  • Posts: 6414
  • Liked: 9104
  • Likes Given: 885
Re: Amazon Project Kuiper Broadband Constellation
« Reply #139 on: 07/11/2020 03:22 am »
https://twitter.com/AjitPaiFCC/status/1281693437658574848

Quote
Earlier today, I shared with my colleagues a proposal to approve, with conditions, #Kuiper’s application to deploy and operate an NGSO satellite network. Satellite constellations like this aim to provide high-speed broadband service to consumers in the U.S. and around the world.

Tags: kuiper 
 

Advertisement NovaTech
Advertisement Northrop Grumman
Advertisement
Advertisement Margaritaville Beach Resort South Padre Island
Advertisement Brady Kenniston
Advertisement NextSpaceflight
Advertisement Nathan Barker Photography
1