Thursday, March 16, 2006 Volen 101, 2:00-3.00 pm
Abstract:
The talk will argue for a "grey-box" reduction notion in distributed computing,
as an alternative to
"black-box" reductions inherited from the classical theory of computing. The
grey-box reduction
classifies problems according to the amount of information about failures
needed to solve them, and
thus captures the main source of "hardness" in distributed systems. This notion
will be illustrated
through atomic object implementations with some rather surprising results.
Host: Liuba Shrira