 
Summary: The Center for Control, Dynamical Systems, and Computation
University of California at Santa Barbara
Spring 2009 Seminar Series
Presents
Linear Iterative Strategies for Information Dissemination
and Processing in Distributed Systems
Shreyas Sundaram
University of Illinois at Urbana  Champaign
Monday, June 1, 2009 10:00  11:00 AM HFH 4164
Abstract: A core requirement in distributed systems and networks is the ability to disseminate information
from some or all of the nodes in the network to the other nodes. In this talk, we describe a linear iterative strat
egy for information dissemination, where each node repeatedly updates its value to be a weighted linear com
bination of its previous value and those of its neighbors. We show that this strategy can be compactly modeled
as a linear dynamical system, and use controltheoretic tools (such as observability theory, structured system
theory, and linear system theory) to characterize its capabilities. First, we show that in connected networks with
timeinvariant topologies, the linear iterative strategy allows every node to obtain the values of all other nodes
after a finite number of iterations (or timesteps). The number of timesteps required is determined by the net
work topology and, in fact, may be minimal over all possible strategies for information dissemination. Next, we
demonstrate the ability of the linear iterative strategy to handle a set of malicious (and possibly coordinated)
nodes that update their values arbitrarily at each timestep. It has been established in the literature that when
