From: Dan Clemmensen (Dan@Clemmensen.ShireNet.com)
Date: Tue Nov 24 1998 - 17:26:22 MST
Ken Meyering wrote:
>
> Dan@Clemmensen.ShireNet.com (Dan Clemmensen):
> > Ken Meyering wrote:
> > > Would it be possible to have a large supercomputer made up of such
> > > nodes, such that the same devices could also be a wireless
> > > telecommunications infrastructure, at the same time as being a
> > > distributed computing system?
> > >
> > > You could make it such that you mass produce the boxes (with
> > > superscalar RAM), to achieve the lowest possible per unit cost.
> > >
> > The comms component will cost a good deal more than the rest of the
> > computer to achieve perhaps 2Mbps connectivity. Cheaper to spend
> > about ten bucks per computer for 100baseT ethernet.
>
> See http://www.wavtrace.com/pressrls081798.htm
>
> Jason Abboud is a cool guy. Very down to earth. He needs help,
> though, creating a strong demand for the product, since he wants to
> achieve excellent economies of scale, to keep costs down.
>
> Maybe we could get him hooked up with Prof. Frank Dietz.
>
> See http://dynamo.ecn.purdue.edu/~hankd/
>
> Maybe we could get these *wireless* devices down to the $50/box
> prices that your 100base-T Ethernet cards are going for today, to
> produce a Linux-based parallel supercomputer that can house the CRIT
> (http://crit.org) frame-level HDTV annotation database and wavelet
> HDTV frame proxies. (Maybe Larry Ellison will help with a shareware
> parallel version of Oracle? It's all a game to these guys, right?
> Set the ego's aside: Bill slips Ted a couple B's to slip to Larry?)
>
I'm very sorry, Ken, but I will not be responding to further posts from you
on this topic. Your references are simply not relevant. The wavtrace stuff is
a point-to-multipoint system intended for metropolitan-area bypass.
The gear almost certainly costs tens to hundreds of thousands of
dollars per site and involves big rooftop antennas. It's cheap
only when compared to running new fiber in a city, and the total bandwidth
is on the order of 100Mbps for all stations. Don't confuse the carrier
frequency with the data rate. So, a 256-node system would permit each node
to generate about .5Mbps (albiet broadcast) at a comms cost on the order of
>10K$/node. The 100BaseT grid permits a total bandwitch of 1.6Gbps with
a per-node contrib of 7Mbps at a current, off-the shelf cost of <50$/node.
furthermore, the multicast radio scheme scales poorly, while the grid
scales well. The grid is only the simplest of a family of doubly-connected
node topologies, and there are obviously higher-order toplologies. The
current $50/node is high compared to purpose-built dual-port NICs
on the motherboard, and 100Mbps is easy to beat, so the off-the-shelf
grid is easy to beat if you care to.
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:49:51 MST