Login
Computer Science @ Rochester
Tuesday, April 19, 2005
9:00 AM
CSB 601
Ph.D. Thesis Proposal
Chengliang Zhang
University of Rochester
Behavior Models for Computer Programs and Virtual Communities
This research develop new models for the behavior of computer programs and virtual communities. For computer programs, we first study the temporal-spatial model of data locality---reference affinity. We discuss the complexity issues of strict reference affinity and show its use in modeling locality hierarchy in divide and conquer programs. We generalize it to weak reference affinity by introducing probability. We propose a sampling method to find an approximate solution. We also study the data placement method for specific hardware configurations. Another problem we consider is dynamic memory management. Using the phase information, we predict the memory usage trend, detect memory leaks, optimize the memory allocator, and design a new preventive garbage collection. For virtual communities, we analyze and predict behaviors for postings and virtual individuals in virtual communities such as slashdot.org.

This proposal defines problems, provides relative background knowledge, summarizes previous work, reports the preliminary results, discusses the future work and the milestones.