1996-06-05 - Re:cycle market

Header Data

From: mccoy@communities.com (Jim McCoy)
To: cypherpunks@toad.com
Message Hash: 442f92b4840deb1edcb1b54118e51bdbed5e465674d9ee5fc494db7426c318d0
Message ID: <v02140b00adda538b7b78@[205.162.51.35]>
Reply To: N/A
UTC Datetime: 1996-06-05 08:23:25 UTC
Raw Date: Wed, 5 Jun 1996 16:23:25 +0800

Raw message

From: mccoy@communities.com (Jim McCoy)
Date: Wed, 5 Jun 1996 16:23:25 +0800
To: cypherpunks@toad.com
Subject: Re:cycle market
Message-ID: <v02140b00adda538b7b78@[205.162.51.35]>
MIME-Version: 1.0
Content-Type: text/plain


Bryce writes:
> I was trying to think what I would use a cycle market for.  The
> only thing that I wait for more than a minute on currently is
> compiling.

Tasks which are inherently parallel (preferably ones where the
various processes are loosly coupled and the need for inter-task
communication is low) are the first things that come to mind,
because advances in the speed of CPUs does not give you the same
advancement increment in solving these problems.  Some examples:

        -rendering and ray tracing
        -simulation
        -analysis of large datasets
        -genetic algorithms and genetic programming (something I am
         working on creating a "cycle market" for in my spare time)

The other class of tasks which could use such a market are those which
are limited more by bandwidth than by processor speed.  If the server can
be given many "heads" which can provide the service (with the back-end
processing done by cycle markets) then it is possible to make significant
gains in distributing the I/O load and getting around network latency by
creating server which is "virtually omnipresent."

jim







Thread