Re: idle CPU markets

[srao@ISI.EDU (Santosh Rao)]

> Date: Fri, 27 Oct 1995 06:25:34 -0700
> To: cypherpunks@toad.com
> From: Jay Campbell <edge@got.net>
> Subject: Re: idle CPU markets
> Sender: owner-cypherpunks@toad.com

> A graphic-arts chum of mine is really hyper about cycle sales - it will
> allow him to produce broadcast-quality commercials on a project-by-project
> basis without a huge initial hardware upgrade investment. Is anyone working
> on a Java libarary/demo/whatnot that allows plug-in algorithms, work
> completion 'fetch next batch' code, and some rudimentary authentication
> scheme? If not, I may tinker with this - if so, I'd like to chat :)


At USC/Information Sciences Institute, we are developing a system that will
enable computer owners to sell or barter processor-cycles on the Internet.
This is in some ways similar to what you talk about, in that it allows
the execution of plug-in applications, and the policies for allocation
and deallocation of remote resources are decided by the owners and are 
transparent to the end user. However, unlike Java, our applications are 
compiled C programs that execute on *compute servers* rather than on
client machines. The application could in turn be an interpreter for 
some other language (e.g. Java). We also support parallel applications 
that use the Parallel Virtual Machine (PVM) libraries for interprocess 
communication. 

We have implemented a prototype known as the Prospero Resource Manager 
(PRM); a parallel version of POVRAY raytracing program is one of the 
applications we've run using PRM. We are currently working on integrating 
PRM with authentication and payment products (including kerberos, netcheque
and netcash) developed as part of our other projects on security and 
electronic commerce. If you are interested in further information about 
any of these products, or are interested in participating in any aspect 
PRM's development we'd be happy to talk to you. More information is also 
available on our web pages: http://nii-server.isi.edu/info/prm and
http://nii-server.isi.edu/gost-group.

We also believe that one of the parallel applications for which PRM is 
useful is showing the weakness of certain encryption methods with limited
key size. If anyone has suggestions for algorithms or software that might
be included as a demo application in the next release of PRM, please contact
me.

Santosh Rao.