From weidai@eskimo.com Wed Dec 17 23:17:14 2003 From: Wei Dai To: cypherpunks-legacy@lists.cpunks.org Subject: idle CPU markets Date: Wed, 17 Dec 2003 23:17:14 +0000 Message-ID: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============5280546805112240252==" --===============5280546805112240252== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit With many high speed personal computers on the Internet and the deployment of low transaction cost Internet payment schemes, it seems inevitable that markets for idle CPU cycles and memory will develop. An interesting problem is to try to predict who this market will benefit, and what the market will be used for. So far it seems that cryptanalytic problems (e.g. factoring and brute forcing of keys) have the highest marginal value/MIPS among problems amenable to loosely coupled distributed computation. However, I think it would be wasteful if the demand in idle CPU and memory markets were to be dominated by cryptanalysts since (non-academic) cryptanalysis is basicly a zero-sum game. When a key is broken, no wealth is created, rather it is transfered from the owner of the key to the cryptanalyst. What other problems would benefit from easy access to lots of distributed CPU cycles? Wei Dai --===============5280546805112240252==-- From wilcoxb@nagina.cs.colorado.edu Wed Dec 17 23:17:14 2003 From: Bryce To: cypherpunks-legacy@lists.cpunks.org Subject: Re: idle CPU markets Date: Wed, 17 Dec 2003 23:17:14 +0000 Message-ID: <199510270342.VAA13430@nagina.cs.colorado.edu> In-Reply-To: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============6131507926863820422==" --===============6131507926863820422== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit -----BEGIN PGP SIGNED MESSAGE----- An entity calling itself "Wei Dai " allegedly wrote: > > What other problems would benefit from easy access to lots of distributed > CPU cycles? Well there are plenty of applications that want lots of cycles, but there are several problems with the idle CPU market approach to getting those cycles. (Examples I can think of: rendering of movie-quality graphics in non-real-time, scientific computation/modelling, compiling...) 1. Many such applications want their computation to be highly responsive-- the long turn-around involved in farming your task over a WAN is often prohibitive. (E.g. real-time graphics.) 2. Also many applications that need this kind of power are highly sensitive to inaccuracy or fraud. A scientific modelling experiment which uses zillions of cycles can be rendered completely worthless if a tiny calculation that had been farmed to Joe Blow is done wrong or is lied about by Joe. 3. Similarly, many such applications are highly confidential. I'm sure some cypherpunks have good ideas on how to deal with problems 3 and maybe 2... Bryce signatures follow "To strive, to seek, to find and not to yield." bryce@colorado.edu -----BEGIN PGP SIGNATURE----- Version: 2.6.2 Comment: Auto-signed under Unix with 'BAP' Easy-PGP v1.01 iQCVAwUBMJBVJvWZSllhfG25AQG5DgP/SL0fcuwTtc140OANZpZZ0jIfiXzyE7/v 0P+vOU1o2sgloAge8drPAo6O3/x92sU1YJFN4QxFLLIpD84vxTjS1XwFAywZqQQu sn8HcNyTVRyUwjgVm3zd4adyQVBorYNpnreDqAnNPMaB9fDHrWMy+09hFx9JV+/X nL15ZctEHTI= =vI1Z -----END PGP SIGNATURE----- --===============6131507926863820422==--