Amazon Project Kuiper Broadband Constellation

[DanClemmensen]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

You need minimum latency, and satellite is the solution? Umm 🤔

LEO satellite has lower latency than terrestrial fiber for some routes. That's because the speed of light in fiber is 200,000 km/s, but the speed of light in vacuum is 300,000 km/s. Also, satellite-to-satellite hops are straight lines while fiber follows crooked rights-of-way, etc.
short-haul: fiber is faster. long-haul, LEO satellite is probably faster.  mid-range: it will vary.  Of course, a fiber net can carry ore data and will therefore potentially have less congestion, but this varies all over the place.

[joek]

...
What unique capability does kuiper bring to the table to address this 'need' that is not already offered by the existing alternatives.

I see kuiper as an equal alternative, I struggle to see any unique capabilities that kuiper brings to the table with regards to AWS.

Depends on how you define "existing alternatives". Terrestrial? LEO? MEO? GEO? All have different capabilities-limitations. Much different and longer conversation, not appropriate for this thread.

With respect to AWS, agree; see no unique capabilities vs. other LEO constellations (or any other transport),. Other than maybe not-SpaceX, or AWS wants to control the transport to a degree alternatives will not provide, or some-such?

One might as easily cast this conversation in terms of AWS laying its own terrestrial fiber or some other transport. They would do that because ...?

[Rebel44]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

I work in telco and I don't know anyone who would care enough to spend this kind of extra $ (not even government agencies) - exceptions already have (much cheaper) dedicated secure physical infrastructure.

[joek]

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
...

Don't see it. First rule of network security: Don't trust the network (transport). If you do, you've lost the game.

[mn]

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
...

Don't see it. First rule of network security: Don't trust the network (transport). If you do, you've lost the game.

Of course you don't trust the transport, that's why you encrypt everything. That works for the vast majority of your potential customer base.

[mn]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

You need minimum latency, and satellite is the solution? Umm 🤔

LEO satellite has lower latency than terrestrial fiber for some routes. That's because the speed of light in fiber is 200,000 km/s, but the speed of light in vacuum is 300,000 km/s. Also, satellite-to-satellite hops are straight lines while fiber follows crooked rights-of-way, etc.
short-haul: fiber is faster. long-haul, LEO satellite is probably faster.  mid-range: it will vary.  Of course, a fiber net can carry ore data and will therefore potentially have less congestion, but this varies all over the place.

I am aware that in theory leo can give you less latency. Does it happen in real life? And what bandwidth can you get for that? The scenario where satellite would be a better solution overall seems to be so narrow?

[DanClemmensen]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

You need minimum latency, and satellite is the solution? Umm 🤔

LEO satellite has lower latency than terrestrial fiber for some routes. That's because the speed of light in fiber is 200,000 km/s, but the speed of light in vacuum is 300,000 km/s. Also, satellite-to-satellite hops are straight lines while fiber follows crooked rights-of-way, etc.
short-haul: fiber is faster. long-haul, LEO satellite is probably faster.  mid-range: it will vary.  Of course, a fiber net can carry ore data and will therefore potentially have less congestion, but this varies all over the place.

I am aware that in theory leo can give you less latency. Does it happen in real life? And what bandwidth can you get for that? The scenario where satellite would be a better solution overall seems to be so narrow?

I have no idea. I gave a purely theoretical response. In theory, as capacity of the satellite system increases and the number of satellites increases, the latency variance will decrease and available capacity increase for the highest-priority traffic. I absolutely agree that the applicability may  be so narrow as to be non-existent. I can think of several ways to use this for internal network control traffic to better optimize the routing of the user traffic, but that's about it. (I have seen quite a few exotic schemes proposed in my 40+ years in networking. Almost none of them were practical.)

[TrevorMonty]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

You need minimum latency, and satellite is the solution? Umm

LEO satellite has lower latency than terrestrial fiber for some routes. That's because the speed of light in fiber is 200,000 km/s, but the speed of light in vacuum is 300,000 km/s. Also, satellite-to-satellite hops are straight lines while fiber follows crooked rights-of-way, etc.
short-haul: fiber is faster. long-haul, LEO satellite is probably faster.  mid-range: it will vary.  Of course, a fiber net can carry ore data and will therefore potentially have less congestion, but this varies all over the place.

Anytime data goes through repeater or internet router there is additional delays that add latency.  It takes time for transistors to turn on and off.

[DanClemmensen]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

You need minimum latency, and satellite is the solution? Umm

LEO satellite has lower latency than terrestrial fiber for some routes. That's because the speed of light in fiber is 200,000 km/s, but the speed of light in vacuum is 300,000 km/s. Also, satellite-to-satellite hops are straight lines while fiber follows crooked rights-of-way, etc.
short-haul: fiber is faster. long-haul, LEO satellite is probably faster.  mid-range: it will vary.  Of course, a fiber net can carry ore data and will therefore potentially have less congestion, but this varies all over the place.

Anytime data goes through repeater or internet router there is additional delays that add latency.  It takes time for transistors to turn on and off.

We have switching delay, queueing delay, transit delay, and transmission delay. Switching delay is the processing from last bit received by a relay until the packet is ready to transmit, and it is affected by those transistors. Queueing delay is the time a packet spends waiting for the packets ahead of it to be transmitted.Transit delay is affected of speed of light and length of the transit. Transmission delay is affected by the encoding and the bit rate is is the time from transmission of the first bit to transmission of the last bit of t4h packet.

* Light in vacuum travels at 300 meters per microsecond.
* a 1000-bit packet has a one microsecond transmission delay at 1 Gbps.
* In the modern world, switching delay can be as low as 1 microsecond. A processor operating a 1 GIPS processes 1000 instructions in one microsecond. Modern relays (routers, switches, satellites) use specialized hardware that pipelines the processing.

So, yes, switching delay is real and back in 1970 when processors operated at 1 MIPS and I entered the business, it mattered a whole lot. Our "high speed trunks" operated at 9600 BPS and transmission delay mattered a whole lot also. Today, both of these are inconsequential in trunking applications, and it's all about queueing delay and transit delay. Transmission delay add the equivalent of 0.2 km of  fiber to the distance. Switching delay adds another 0.2 Km of fiber.

Note that in my original reply I specified high-priority traffic. This stuff goes to the head of each queue as it transits the network, and the amount must be kept so small that the high-priority queue at each node must (almost) always be empty.

[TrevorMonty]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

You need minimum latency, and satellite is the solution? Umm

LEO satellite has lower latency than terrestrial fiber for some routes. That's because the speed of light in fiber is 200,000 km/s, but the speed of light in vacuum is 300,000 km/s. Also, satellite-to-satellite hops are straight lines while fiber follows crooked rights-of-way, etc.
short-haul: fiber is faster. long-haul, LEO satellite is probably faster.  mid-range: it will vary.  Of course, a fiber net can carry ore data and will therefore potentially have less congestion, but this varies all over the place.

Anytime data goes through repeater or internet router there is additional delays that add latency.  It takes time for transistors to turn on and off.

We have switching delay, queueing delay, transit delay, and transmission delay. Switching delay is the processing from last bit received by a relay until the packet is ready to transmit, and it is affected by those transistors. Queueing delay is the time a packet spends waiting for the packets ahead of it to be transmitted.Transit delay is affected of speed of light and length of the transit. Transmission delay is affected by the encoding and the bit rate is is the time from transmission of the first bit to transmission of the last bit of t4h packet.

* Light in vacuum travels at 300 meters per microsecond.
* a 1000-bit packet has a one microsecond transmission delay at 1 Gbps.
* In the modern world, switching delay can be as low as 1 microsecond. A processor operating a 1 GIPS processes 1000 instructions in one microsecond. Modern relays (routers, switches, satellites) use specialized hardware that pipelines the processing.

So, yes, switching delay is real and back in 1970 when processors operated at 1 MIPS and I entered the business, it mattered a whole lot. Our "high speed trunks" operated at 9600 BPS and transmission delay mattered a whole lot also. Today, both of these are inconsequential in trunking applications, and it's all about queueing delay and transit delay. Transmission delay add the equivalent of 0.2 km of  fiber to the distance. Switching delay adds another 0.2 Km of fiber.

Note that in my original reply I specified high-priority traffic. This stuff goes to the head of each queue as it transits the network, and the amount must be kept so small that the high-priority queue at each node must (almost) always be empty.

Who decides if the data is high priority when going through international routers.

[DanClemmensen]

I don't see the benefit. Data in transport is already secured by encryption.

I can see the value for some customers. The very act of communicating, its quantity and timing, says something.
And Quantum is real and growing. If you need maximum security and minimum latency, satellite makes sense.

You need minimum latency, and satellite is the solution? Umm

LEO satellite has lower latency than terrestrial fiber for some routes. That's because the speed of light in fiber is 200,000 km/s, but the speed of light in vacuum is 300,000 km/s. Also, satellite-to-satellite hops are straight lines while fiber follows crooked rights-of-way, etc.
short-haul: fiber is faster. long-haul, LEO satellite is probably faster.  mid-range: it will vary.  Of course, a fiber net can carry ore data and will therefore potentially have less congestion, but this varies all over the place.

Anytime data goes through repeater or internet router there is additional delays that add latency.  It takes time for transistors to turn on and off.

We have switching delay, queueing delay, transit delay, and transmission delay. Switching delay is the processing from last bit received by a relay until the packet is ready to transmit, and it is affected by those transistors. Queueing delay is the time a packet spends waiting for the packets ahead of it to be transmitted.Transit delay is affected of speed of light and length of the transit. Transmission delay is affected by the encoding and the bit rate is is the time from transmission of the first bit to transmission of the last bit of t4h packet.

* Light in vacuum travels at 300 meters per microsecond.
* a 1000-bit packet has a one microsecond transmission delay at 1 Gbps.
* In the modern world, switching delay can be as low as 1 microsecond. A processor operating a 1 GIPS processes 1000 instructions in one microsecond. Modern relays (routers, switches, satellites) use specialized hardware that pipelines the processing.

So, yes, switching delay is real and back in 1970 when processors operated at 1 MIPS and I entered the business, it mattered a whole lot. Our "high speed trunks" operated at 9600 BPS and transmission delay mattered a whole lot also. Today, both of these are inconsequential in trunking applications, and it's all about queueing delay and transit delay. Transmission delay add the equivalent of 0.2 km of  fiber to the distance. Switching delay adds another 0.2 Km of fiber.

Note that in my original reply I specified high-priority traffic. This stuff goes to the head of each queue as it transits the network, and the amount must be kept so small that the high-priority queue at each node must (almost) always be empty.

Who decides if the data is high priority when going through international routers.

The very highest priorities are internal and carry only traffic control packets: these are generally at layers below the IP layer. QoS at the IP layer is negotiated by the carriers. It's complicated. My explanation was intended to be basic and applies to the lower layers in all networks. As you seem to imply, yes, if you want absolute control of QoS, you must own the network. This has been true since the invention of the telegraph and has not changed.

The original discussion was about security, not QoS. You can layer security atop an insecure network, and in fact it is crazy to trust the network even if you own it.

[joek]

Who decides if the data is high priority when going through international routers.

Not sure what you mean by "international routers"? In any case, the carrier (transport owner) ultimately decides as they control the transport. However, as with any such arrangements, depends on a number of factors including contractual stipulations, peering arrangements, etc.

Don't think we want to go down that rabbit hole in this thread, otherwise we will be here forever discussing how the Internet works. Having grown up with it (an IMP in the machine room and later a fuzzball on my desk), can only say this is not time or place to discuss. Plenty of resources out there for the curious.

[TrevorMonty]

If data is going from one side of world to other then packets will pass through multiple international routers/exchanges.

[Nomadd]

 My experience is pretty much telegraph era, but I remember that qos could be quite an issue over multiple networks. Early on, you needed it to keep voice buffers small for good service, and later on for stock traders to have a few milliseconds advantage over their neighbors. Cooperation between pipelines wasn't always smooth.
 I always thought that point to point would be a good market for leos, without data touching the ground or outside networks between points.

[DanClemmensen]

If data is going from one side of world to other then packets will pass through multiple international routers/exchanges.

This is a gross oversimplification and is not true in many cases (e.g., new York to Hawaii). Please be more specific, and please state how your underlying point relates to the topic, which is "Amazon Project Kuiper Broadband Constellation".

[TrevorMonty]

If data is going from one side of world to other then packets will pass through multiple international routers/exchanges.

This is a gross oversimplification and is not true in many cases (e.g., new York to Hawaii). Please be more specific, and please state how your underlying point relates to the topic, which is "Amazon Project Kuiper Broadband Constellation".

Try getting data from Taiwan to Poland (AWS locations) as an example as you are struggling to think outside USA.

[DanClemmensen]

If data is going from one side of world to other then packets will pass through multiple international routers/exchanges.

This is a gross oversimplification and is not true in many cases (e.g., new York to Hawaii). Please be more specific, and please state how your underlying point relates to the topic, which is "Amazon Project Kuiper Broadband Constellation".

Try getting data from Taiwan to Poland (AWS locations) as an example as you are struggling to think outside USA.

I still do not understand your point. I have no problem with thinking outside the USA, I was merely providing a counterexample. I infer from your reply that you feel Amazon will have more control over the data that flows strictly within the Kuiper constellation that is does of traffic that flows along a multi-AS path. Is this correct?

[meekGee]

I don't think there's a doubt that some data can benefit from a super low latency pipe, even if the bandwidth is limited. Some of that data will go to/from data centers.


There's also data from dispersed locations where land connections are limited.

I'm also reasonably confident that land routes will generally have more capacity than wireless ones, especially data center to data center.

There's no black and white here. Starlink/Kuiper will have markets.  What's not clear is why AWS can't use Starlink.  We're talking years in time to market, a much larger constellation, and two generations ahead.  EDIT: oh, and cheaper to launch.  By a lot.

It's not like Starlink will engage in corporate espionage of customers that happen to  work with AWS.

[RedLineTrain]

To be clear, above I was being a little bit of a devil's advocate.  I doubt the business case for Kuiper somewhat, but like meekGee, I don't think this is black and white.  Certainly, Amazon will find uses for Kuiper, if it doesn't pull the plug on the whole project.

However, I'm guessing that Amazon got a little bit in over its skis.  Starlink has become a much more ambitious proposition than Musk laid out in 2015 and was apparent in 2018, when Kuiper was initiated.  The barriers to Kuiper's thriving are much higher now than they were (they are now eyewatering), not least because Starlink Gen1 and Gen2 actually exist now and it is signing up customers hand-over-fist.

One aspect missing in our discussion is that Kuiper is additive to the licensed bandwidth supply.  Starlink is NCo=1, meaning that it can supply two overlapping beams in a frequency, one for each network:  Gen1 and Gen2.  Kuiper I believe is also NCo=1, meaning that it can overlap once with Starlink's frequencies.  In a sense, Kuiper's unique service is simply that it isn't Starlink.  I don't know how to value that uniqueness, however.

[joek]

If data is going from one side of world to other then packets will pass through multiple international routers/exchanges.

Let's clarify that: "... packets will pass through multiple international routers/exchanges owned and operated by different companies-governments-collectives-whatever". That's the way the Internet works. In that context, Kuiper will be just another carrier among many. Kuiper may provide something special to their subscribers, or on behalf of there subscribers. So what is that "something special"?