home · mobile · calendar · defenses · 2005-2006 · 

Thesis Defense - Crawl

Affinity-Directed Mobility
Daniel Crawl
Computer Science PhD Candidate

As computer users work on an increasingly diverse array of mobile computing devices, fragmentation (portions of related data located on many separate devices) and versioning (different versions of the same datum located on different devices) of user data have become increasingly prevalent and important problems. The contributions of this dissertation are: (1) the design and implementation of a novel mechanism for addressing fragmentation and versioning using affinity relationships; (2) the evaluation of this mechanism through trace-based simulation; and (3) the design, implementation, and evaluation of a modular architecture to process file system events that underlies this mechanism, but which has broader application.

Affinity relationships constructed at runtime help identify data of current interest to users; these relationships then guide the caching of user data on mobile computing devices. Trace data collected over a 15 month period has been used to inform the design of the affinity mechanism. Trace-based simulation demonstrates the effectiveness of an affinity-directed approach in reducing fragmentation and versioning while incurring low overhead.

This dissertation describes the design, implementation and evaluation of an abstraction layer and framework that permits the development of portable, distributed applications that run efficiently on a variety of operating systems. The modular design allows users to implement complex processing tasks by chaining a handful of simple modules. File system traces were used to demonstrate a 90% performance improvement, on average, when using this method, compared to an application that periodically scans the file system each day. Further gains can be achieved with the introduction of hysteresis.

Committee: John Bennett, Professor (Chair)
Kenneth Anderson, Associate Professor
Dirk Grunwald, Associate Professor
Richard Han, Assistant Professor
Evan Speight, IBM Corporation
Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:20)