[Beowulf] [Server-sky] Server Sky - Internet and computation in orbit

Eugen Leitl eugen at leitl.org
Fri Jun 21 13:26:28 PDT 2013

----- Forwarded message from Keith Lofstrom <keithl at gate.kl-ic.com> -----

Date: Fri, 21 Jun 2013 13:01:22 -0700
From: Keith Lofstrom <keithl at gate.kl-ic.com>
To: Eugen Leitl <eugen at leitl.org>
Cc: server-sky at server-sky.com, "Lux, Jim (337C)" <james.p.lux at jpl.nasa.gov>
Subject: Re: [Server-sky] [Beowulf] Server Sky - Internet and computation in orbit
Reply-To: keithl at keithl.com
User-Agent: Mutt/

Most questions are answered on the wiki, http://server-sky.com 
There's a search function that will help sift through the hundreds of pages.
Plenty of unanswered questions, but that is why I am sharing the idea,
so others can contribute and claim some of the credit.

Forwarded message from "Lux, Jim (337C)" <james.p.lux at jpl.nasa.gov>
> 1) orbital debris - fling those thousands of widgets out there.  Are they high enough to stay in orbit for a while? Are they going to damage things that hit them? 

6411km altitude / 12789km radius, about one Re out.  Thinsats might
damage some objects in crossing orbits if we let collisions happen,
but arrays are capable of both continuous maneuvering and high
luminance lookdown radar.  There aren't many objects in crossing
orbits, and all are derelicts, according to the NORAD database.

Thinsats are much thinner than the Whipple shields enclosing
most critical spacecraft systems.  They will remove some paint
and some material underneath, but they are unlikely to do
anything more than micrometeoroids do.

And your next question will be about radiation - we are aiming
for the lower van Allen belt.  I'm a chip designer.  Recent
rad-hard developments in semiconductors and integrated circuits
are what led me to server-sky.  Look at the website for more.

> 2) orbital mechanics - the "array" pretty much has to be flat, that is, they're all at the same orbit height, otherwise they'll drift apart, since the period is different.   

All thinsats in an array have identical orbital periods (and
the same semimajor axis, to first order), though all have
slightly different orbital elements.  The arrays rotate,
though their shape is skewed by the velocity changes associated
with radius variations.  Rule of thumb - a 1 meter radial
"apogee" turns into a 2 meter retrograde displacement relatve
to array center 1/4 orbit later.  To see how an array of thinsats
evolves over an orbit, look at http://server-sky.com/IEEESustech2013

There are small second order distortions (J₂ and light pressure,
for example), factored into array shape and central orbit choice,
and third order distortions ( lunar/solar/jupiter tides for example )
that will be dealt with by orbit shape again.  The small residual
errors ( < 1E-9 of orbit velocity ) can be dealt with by light sail

> 3) does it really save anything to put the computation in orbit?

Server Sky doesn't save much for developed nations.  Google and Amazon
will still build 100 megawatt data centers and pump CO₂ into the
atmosphere while pretending to be "green".  Much of the CO₂ generated
while making the concrete and metal and plastic constituting 99.99%
by weight of those data centers.  Part of the pretense is building
solar arrays as if replacing plants with solar panels and destabilizing
the electrical grid with intermittent and unpredictable power demand
somehow helps the environment.

It is quite different for poor nations with large rural populations
that haven't deployed much communication infrastructure.  India has
400,000 cell towers, connected by microwave and powered by diesel,
with the "backbone" being a mix of fiber and microwave that follows
their rail network.  No way they can afford to deploy much fiber
beyond that.   Even their microwave grid is way oversubscribed with
the rapid uptake of cell phones.  Bringing broadband to half a
billion rural Indians can create trillions of dollars a year of 
economic value.  The cheapest way to do that is with computation,
and large-aperture/millimeter-wave communication in orbit.

> I'd like to see more justification of the 100x cost differential
> between ground and space 25 years from now. 

My career as a chip designer has spanned a 1 billion times decrease
in the cost of a transistor, and a 1 million times increase in the
speed-power capability of transistors.  By 2015, we will have deployed
a zettatransistor, 1e21 devices.  Much of that growth came from 
invading and transforming seemingly unrelated fields, like biology
and publishing and transportation.  We point the transistor hose at
problems and wash them away.

Meanwhile, space development has stagnated - we've built a bunch
of plausible-but-wrong systems like shuttle, but our workhorse
launchers (Atlas and Proton) are incremental improvements of
1960 designs.  The major advances in space technology have
been electronic systems like Opportunity, using semiconductors
10 years behind the stuff you can buy at Walmart, and system
construction techniques resembling 1950s aircraft.

Craig Venter used modern semiconductor technology to sequence the
human genome, turning a planned two decade multibillion dollar 
government effort into a 2 year 300 million dollar transformation 
of biology.  As a direct result of his work, you can get a 1M-SNP
genotyping for $99, and can expect a whole genome sequence for
less than $1K in a year or two.  

I've spent my whole life waiting for the "cheap-rocket-first" 
community to deliver.  I'm not waiting any more.  I'm not nearly
as smart as Craig Venter, but shifting attention from rockets to
electronics and market needs doesn't require the same brainpower,
just assembling off-the-shelf processes into new systems serving
ignored markets, and doing so at Moore's law rates.   There's a
cornucopia of new technology and new opportunities out there,
and those who cling to the old ideas will be left in the dust.

25 years?  I've organized chip products in 25 weeks (design/fab/
test/sell).  Server sky's long lead times involve team building at
the front end, and negotiating with the ITU at the back end.  There
are plenty of revenue opportunities along the path to deployed
arrays, and a team whose combined cleverness far exceeds mine will
be rich before we put the first thinsat in space.  The job now is
to define viable long term goals, discover people who share them,
and develop a search process that finds the easiest and most
lucrative stepping stones on the shortest path from here to there.

There are certainly trillions of plausible-but-wrong ways to do
this, and with bad planning my team will pick one of those. 
But the opportunity is far too large to be ignored, and even a
magnificent failure will inspire thousands of others who think
they can do better.  Some of them will be right.


Keith Lofstrom          keithl at keithl.com         Voice (503)-520-1993

----- End forwarded message -----
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
ICBM: 48.07100, 11.36820 http://ativel.com http://postbiota.org
AC894EC5: 38A5 5F46 A4FF 59B8 336B  47EE F46E 3489 AC89 4EC5

More information about the Beowulf mailing list