[Colloq] PhD Thesis Defense by Dan Kunkle, Thursday, Dec. 16

[Rachel Kalweit]

The College of Computer and Information Science presents:

PhD Thesis Defense by:
Daniel Kunkle

Date: Thursday, December 16, 2010
Time: 12:00 noon
Location: 166 West Village H

Title: "Roomy: A New Approach to Parallel Disk-based Computation"

Abstract:

Disk-based computation provides a major new application of disks in addition to
the three traditional uses: file systems, databases, and virtual memory.
Disk-based computation uses the many disks already available in a computational
cluster as main memory. In doing so, disk is elevated to a level normally
reserved for RAM, and applications are provided with several orders of
magnitude more working space for the same price.

Unfortunately, developing efficient parallel disk-based applications is often a
difficult task. To solve this problem, this talk introduces Roomy, a new
programming language extension and open source library for writing parallel
disk-based applications.

The core of this research is in developing a library of general algorithms and
larger applications on top of Roomy. Applications considered include: a proof
that 26 moves suffice for Rubik's Cube; general breadth-first search
algorithms; and a package for very large binary decision diagrams.

In developing these applications, the two central questions addressed are:

 1. What is the class of applications for which parallel disk-based computing
 is practical?

 2. How can existing sequential algorithms and software be adapted to take
 advantage of parallel disk-based computing?

Committee:

Gene Cooperman (advisor)
Pete Manolios
Mirek Riedewald
Fan Yang (Google)