Tom Mueller interview 02 May 2017

[IainMcClatchie]

Caching movies and video makes a lot of sense.  Netflix and Youtube generate a significant fraction of total US and I think worldwide traffic, and are responsible for an even greater fraction during usage peaks.

The amount of storage and power needed to cache movies and video is fairly minor.  10 TB of storage should do it.  In NVMe form that's 20 packages that will fit on a 4 x 5 inch circuit board, cost under $7000, burn 100 watts, and deliver 600 gigabits/second.

I can see tripling the part count to get the flash to last longer, but this won't increase power consumption much.

I don't see caching making much sense for general web stuff... the storage requirements get fairly large.

[launchwatcher]

The amount of storage and power needed to cache movies and video is fairly minor.  10 TB of storage should do it. 

I think you underestimate how much video there is out there.

Netflix is, today, deploying caching appliances into ISP's with significantly larger capacity (up to 280TB).. 
Now consider how much more video's on Youtube..

[RonM]

The latency for SpaceX satellites will be low. No need for server farms in orbit. It's a terrible environment for data storage and pretty hard for the hardware techs to access. Without being able to swap out failed components, capacity will rapidly drop. Just keep data storage on the ground.

[Nomadd]

The latency for SpaceX satellites will be low. No need for server farms in orbit. It's a terrible environment for data storage and pretty hard for the hardware techs to access. Without being able to swap out failed components, capacity will rapidly drop. Just keep data storage on the ground.

If 70% of your uphill bandwidth can be eliminated by storing the most popular movies, Dancing with the Stars episodes and YouTube cute animal videos, it's not a small thing. Seagate just came out with a 60TB, 2.5" SSD that uses 15 watts.
 (The first hard drive I ever maintained was 14MB and the size of a dishwasher) Used on a system with core memory)

[oldAtlas_Eguy]

Too much specific that belongs on the SpaceX a sat vendor topic for this thread.

Although I find it interesting much of this has been discussed in great detail on that other thread as much as a year ago.

[rpapo]

(The first hard drive I ever maintained was 14MB and the size of a dishwasher) Used on a system with core memory)

Good old Memorex 660...

[feynmanrules]

Selling something for more than what it costs is the idea.

Marginal cost in this context don't include depreciation or amortization.    Those do show up on financial statements as part of Cost of Goods Sold (and elsewhere).   As an example tsla has gross margins ((sales - cogs)/sales) between 25-30%, which means if average sale price is $80k then they're "spending" around $60k to make a car. This marginal cost is a more granular # that doesn't include amortization of development costs or depreciation on the plant.    The granularity made it interesting and also that it was 5 years ago (costs tend to drop with scale and learning how better to make one's product).   

Back on topic- Unlike tesla spx is a private company and Mueller didn't disclose a similar # for spacex.   He suggested each merlin 1d is some fraction of a million and suggested 600k as an example.    If that's correct they're spending ~6m on engines.   Perhaps this has been disclosed before but I wasn't aware.

I believe he said marginal cost for a Model S. That means, given their existing physical plant, etc, what is the cost for Tesla to make a Model S vs not making one. That should be much lower than the sale price, as it only includes marginal labor, materials, and maybe wear and tear on production equipment.

[feynmanrules]

Here's the link into the original discussion of this on the satellite thread: http://forum.nasaspaceflight.com/index.php?topic=36552.msg1325507;topicseen#msg1325507

Briefly discussed but robotbeat gets credit for being ahead.

Too much specific that belongs on the SpaceX a sat vendor topic for this thread.

Although I find it interesting much of this has been discussed in great detail on that other thread as much as a year ago.

[Lar]

What part of not talking about fusion on this thread were we having trouble with?  thanks.

Edit: I've splitmerged some posts... there were so many it was a bit of work.... If you think I missed one, PM me or report to mod.

http://forum.nasaspaceflight.com/index.php?topic=5367.0 is where they are now.

[QuantumG]

What part of not talking about fusion on this thread were we having trouble with?  thanks.

Can we talk about how solar panels don't last forever?

Seems like it might be important to what Tom was saying about needing to find other ways to power a Mars base.

[envy887]

What part of not talking about fusion on this thread were we having trouble with?  thanks.

For those wondering, the fusion thread is here:
https://forum.nasaspaceflight.com/index.php?topic=5367.msg1675959#new

[meekGee]

The latency for SpaceX satellites will be low. No need for server farms in orbit. It's a terrible environment for data storage and pretty hard for the hardware techs to access. Without being able to swap out failed components, capacity will rapidly drop. Just keep data storage on the ground.

Also key to caching is locality.  You cache Bollywood movies in India, but not in Siberia.   And vice-versa, whatever that may be.

Cache nearby, and use the constellation for distribution, since "nearby" in practical terms can still be thousands of miles.

[RDMM2081]

The latency for SpaceX satellites will be low. No need for server farms in orbit. It's a terrible environment for data storage and pretty hard for the hardware techs to access. Without being able to swap out failed components, capacity will rapidly drop. Just keep data storage on the ground.

Also key to caching is locality.  You cache Bollywood movies in India, but not in Siberia.   And vice-versa, whatever that may be.

Cache nearby, and use the constellation for distribution, since "nearby" in practical terms can still be thousands of miles.

The caching options on a constellation as large as CommX, with the levels of interconnectedness that mesh/topology might present is quite astounding, and I frankly have no idea where to begin for an analysis or analogy.

[Robotbeat]

We've had this caching conversation I think multiple times before, and it's off topic.

[MP99]

One benefit of being a LEO constellation is that like the ISS, one is under the inner Van Allen belts unlike geosats, so radiation isn't as much a consideration.

Isn't the constellation going to be just in the lower reaches of the belt at 1100 km?

Cheers, Martin

Sent from my Nexus 6 using Tapatalk

[MP99]

Yeah, the idea that you could build a gas generator cycle rocket for less than $10M was once unbelievable. Any day now someone will produce one in their garage, although electric pumps have essentially taken that scale in a different direction.

I once saw a comment that rl10 (not a GG, of course), could be built for about the same as a helicopter engine ($40k?) after a lot of re-engineering.

Cheers, Martin

[Space Ghost 1962]

One benefit of being a LEO constellation is that like the ISS, one is under the inner Van Allen belts unlike geosats, so radiation isn't as much a consideration.

Isn't the constellation going to be just in the lower reaches of the belt at 1100 km?

More complex.

The belts don't have "fixed" properties, but vary in disposition around the earth, and in height given solar activity.

The inner belt can dip below 200km in places, and its usually above 1000km near the equator. Also, all bets are off, even for the ISS, if one were to have an event like in the 1800's, where likely the belts made it into the upper stratosphere ... there were perpetual auroras and sky glow.

Since LEO constellations are collections of near polar planes, it is impossible to orbit under all of them, all of the time. However, geosynchronous sats spend 100% of their time in the upper Van Allen belts, by comparison. And LEO near polar have brief encounters with anomalous extensions like the South Atlantic Anomaly, because they transit them in a fraction of their orbit. Some constellations can be even designed to avoid them given intentionally limited coverage (you don't place a plane there).

Am also not sure how high SX LEO communication sat constellation might really be. They might intentionally have the first ones low, intending them to be short lived, knowing that a second generation would be on the heels of the first.

[MP99]

Many thanks - I learned something.

Cheers, Martin

[docmordrid]

>
Am also not sure how high SX LEO communication sat constellation might really be. They might intentionally have the first ones low, intending them to be short lived, knowing that a second generation would be on the heels of the first.

First constellation table in image.

LEO & VLEO technical PDF's attached.

[DOCinCT]

(The first hard drive I ever maintained was 14MB and the size of a dishwasher) Used on a system with core memory)

Good old Memorex 660...

Back in 1976 the hard drive on the computer in the Pychoacoustic Lab at SUNY Binghamton was either 5 or 10 MB.  Not really the good old days.