Bernardo Huberman: "Networked Computation as Economics"
Lecture to be held on Monday May 22, 2006, 18:15.
Room D 16.2, Main Building, ETH Zurich.
We often use computers to study economic systems and networks, but few people realize that we can use economics to study and design computational systems. The reason is that such systems are characterized by contention for resources by strategic agents which operate in a world with imperfect and delayed information.
In order to illustrate how economics helps the design of networked computers I will present solutions to two problems that plague these systems. The first one is the fair and swift allocation of computational resources to those who need them, and the second is the ability to adaptively deliver specific content to any user regardless of its popularity. We have solved these problems by building Tycoon, a market-based distributed resource allocation system based on an auction share scheduling algorithm and truth-telling reservations. We have also designed an efficient mechanism for peer-to-peer systems that can generate a wide diversity of content through a parimutuel market similar to that commonly used for betting in horse races.