[e2e] The Intra-Planetary Internet
Jonathan M. Smith
jms at central.cis.upenn.edu
Thu Aug 2 05:45:49 PDT 2001
At 08:11 AM 8/2/2001 +0100, Jon Crowcroft wrote:
>for those attending the ietf, there's an ipnrg interest meeting in
>parallel on the thursday (see attached message)
>meanwhile, I think in this era of globalization, we should be paying
>attention to the democratic rights of the planet, and building an
>intra-planetary internet -
>we need to urgently and boldly, go where IP has not gove before:
>straight downwards. We need to figure out how to modulate a signal so
>that instead of going 25,000 miles around the planet, we go 4000 miles
>through it:- i know Order(2 pi) reduction is not great in terms of
>scaling, but its hideously important in terms of latency at the speed
>of light -
>My suggestion is that we look at modulated streams of neutrinos (I'd
>suggest graviyt waves, but people might get heavy about that).
DARPA will have to start a program to work on the VLSI-neutrino transceiver
systems. Maybe they already have it in place... :-;
>We could then realize the Cray dream of a global grid girdling the
>planet, with a maximum delay of 21.5 msecs between the furthest
>computers (or voip devices)....i think this needs urgent attention -
>the spinoffs could be excelletn -even leading into orbit and
>further, since IP over Neutrinos is immune from sun spot and other RF
>Why is The Global GRID like a 3G network ?
"won't work" seems like a simpler answer...
>The GRID is being hyped, much as 3G is hyped - this is to generate a
>lot of investment. It is just possible that it is out of the same
>fear of receession that fuels hiring people to dig ditches then fill
>them in, or to start an arms race. just like 3G , its not obvious that
>there is any business case, or any business at all...It is, as they
>say in the film business, a "stunt project", rather than the real
Look - we've had this happen before. I'm going to bang on the GRID for
a minute. Please excuse the rant (yes, I know it's one). There are three
bloodlines in the CS discipline: math/logic, EE, and numerical
Many years ago, the numerical analysis folks were very important - since
they were solving the problems that governments, who could afford computers,
cared about. They are less important now, as Moore's law has largely
out of existence. Nonetheless, in government bureaucracies, where "big
science" has good demos and can get the attention of politicians who have
"getting" ideas like open source, this breed lives on. The GRID is their
>Continuing the analogy:
>Hot topics in the global grid center around
>programming models and resource brokering.
>The Grid Resource Information Service gathers and , and tfces to
>monitoring systems, and logs that information in a directory, and
>then the various GRID middleware platforms allow you to place a task
>on one or more computing systems according to Processing, Storage,
>interaction and visualisation requirements (subject to any furthre
>special system contraints, and accounting and security limitations).
>3G handsets allow wireless access to information resources, and have
>reasoanble local processing and display facilities, but are short of
>power. Battery life is short and not getting better as fast as Moore's
>law, and the effect of the law on programmers' expectation of
>extra cycles, memory and storage law is consuming it. Nor are users
>good at remembering to re-charge hand held devices predictably (even
>when they try, airport lounges and even business class seats often
>deny them the right).
>Most of the teraflop machines in California are not only short of
>power, they have unreliable power supplies - it is not clear that a 3
>week computation can be carried out uninterrupted - it is clear that
>the typical Cray T3D and IBM SP and Beowolf cluster in the North
>American area would be far better placed in the 3rd world (i.e.
>Europe) where the power supply is reliable. American
>supre-computer/GRID customers would then rely only on trans-atlantic
>network relaibility, which is excellent.
>But what of latency (I hear you cry)? Well, that is a research
>problem that faces both the 2.5G and 3G networks too. With the
>very poor packet access times we find on the so-called broadband
>access links already, being significantly higher than the measured IP
>packet delays on the UK-US links for example, it seems that this is
>really a common ground.
>But supposing we come up with a "best effort" computing model, where
>node failures are no more burdensome that packet loss is in the
>Internet. Then we could put the Big Jobs back in California with
>So then, what about a cominbation of the two? The available cycles on any
>given 3G handset may be poor, but imagine a beowulf cluster with
>as many 1GHz processors as there are GSM handsets today - a tad faster
>than 35,000 teraflops would be avaialble in the UK alone.
>But with a MTBF of 1 hour, and a MTTR of 1 hour, the pgroamming model
>would be a challenge.
>What a prospect...for the real grand challenge.
>what is more, the californians could offset the latency against the
>Kyoto agreed CO2 savings....instead of (or as well as) growing more
>trees. In fact, if the typical 3G handset case was made of wood, we
>would create a complete constructive eco-friendly, pervasive,
>unbiquitous, user-engaged, global democratic decentralised grid.
>I propose that the proper study of the GRID, is:
>Grid Resourcing with Energy Efficiency thru Networking - GREEN.
>As part of the Interplanetary Internet (IPN) research effort, we
>are actively seeking out peer review comments regarding the
>architecture currently defined within:
>In addition to electronic correspondence, whenever possible, we have
>been holding small gatherings for live interaction. We've found such
>meetings to be very useful. An added nice feature of these live gatherings
>is that they allow for more in depth exploration of your interest in more
>direct participation in the IPN effort itself. The upcoming London IETF
>presents such a window of opportunity for such a gathering.
>We would like to solicit your comments/inputs on the current IPN
>architecture and wish to invite you attend to an informal peer review
>gathering the week of the upcoming IETF.
>On Thursday, August 9th Logica is generously providing some meeting
>space at their London facilities to support an informal IPN peer
>review gathering. Because of the overlap with Thursday's IETF sessions,
>we are entertaining the notion of holding 2 separate "opportunistic
>(morning & afternoon) at Logica as well as the possibility of third
>to be held over lunch at a local pub or cafe.
>In the very likely event that WG or other IETF related conflicts prevent
>you from joining us at Logica, we would welcome the opportunity to chat at
>the IETF hotel at a more convenient time during the week.
>If your schedule and your interests accommodate joining us, please let us
>know - including whether or not you are interested in joining us during
>the morning, afternoon or lunch (or any combination).
>If you'd rather chat during a more convenient time (including other than
>IETF week), please let us know that too. Electronic comments are always
>travis at gst.com
>The details of the meeting location are as follows:
>Office : United Kingdom House
>Address : United Kingdom House (5th Floor, Rooms 4 & 5)
> 180 Oxford Street
>Phone : +44 (0)20 7446 3306
>Fax : +44 (0)20 7446 4940
>Host Contacts: Martin Symonds <SymondsM at logica.com>
> Gavin Kenny <KennyGA at logica.com>
>Directions : Turn right out of Oxford Circus tube, (Exit 1) into Oxford
> UK House entrance is on Great Titchfield St, the 2nd
> on the left
More information about the end2end-interest