CS 5944 Speaker Schedule
- Jan 30, 2009
Speaker: Dr. Wu-chun Feng;
Slides, and
Survey Link
Green Supercomputing comes of Age
Abstract: When Green Destiny, a highly energy-efficient supercomputer, debuted
in early 2002, the applications community embraced the solution as
being revolutionary. In contrast, the systems community in
high-performance computing (HPC) greeted Green Destiny with ridicule
and scorn despite having squeezed a 240-node cluster into five square
feet and a thermal power envelope of only 3.2 kW (i.e., two
hairdryers). Green Destiny provided reliable supercomputing cycles
while sitting in an 85-degree F dusty warehouse at 7,400 square feet
above sea level, and it did so without any special facilities, i.e.,
no cooling, no humidification control, no air filtration, and no
ventilation.
In the five years since, power and cooling have finally become
first-class design constraints in HPC due to their effect on total
cost of ownership as well as efficiency, reliability, and availability
(ERA). Thus, this talk will present the evolution of Green Destiny
from an architecturally based low-power approach to a software-based
approach that runs on commodity processors. The talk will then
reflect on how such software based techniques might be pushed back
down into hardware.
- Feb 6, 2009
Speaker: Dr. Alexey Onufriev;
Slides, and Survey Link
The Computational Core of Molecular Modeling
Abstract: The ability to model biological macromolecules at atomic resolution is critical for progress in many areas of biology. What does it take to understand or modify the function of a molecule if its atomic structure is known? What does it take to predict the structure if it is unknown?
In this talk I will describe some of the currently used modeling methods with the focus on computational challenges they present and their possible solutions. These challenges will be discussed in the context of the following specific problems: (1) computer-aided (rational) drug design, (2) understanding of the connection between molecular function and its motion, and (3) the grand challenge of computational science: "the protein folding problem".
- Feb 13, 2009
Speaker: Dr. Manuel Perez;
Slides, and Survey Link
Personal Information Management: "Where is my next meeting?"
Abstract:
With the increasing proliferation of computing-enabled devices and the
resulting overload of digital information, personal information
management (PIM) has become a challenge to all. PIM research is mostly
concerned with studying how people find, keep, organize, and re-find
(or reuse) information in and around their personal information space
(computers, phones, home, desks, etc.). In this presentation, I will
present a brief overview of my group's PIM research focused on trying
to understand the use of multiple devices to manage one's digital
life. The talk will conclude with the implications of some of these
findings on the design of future PIM tools and a view into the future
of PIM research.
- Feb 20, 2009
Speaker: Dr. Steve Harrison;
Slides, and
Survey Link
Creative Computing: The Art of Innovation
Abstract: When we think about technological innovation, we often think first of
the scientist, engineer, or driven tinkerer addressing a problem of great
import, turning what was `science fiction' into technological fact.
There are other ways that innovation occurs; in this talk, I will argue
that one way is the intersection of art and engineering. I'll give some
historical precedents, show a few projects that I have been involved in, and
explain how this approach can get computer scientists to think differently.
Along the way, we'll revisit a few milestones in innovation and the
underlying ideas of human-computer interaction.
- Feb 27, 2009
Speaker: Dr. Caitlin Kelleher; Slides, and Survey Link
Looking Glass: Supporting Learning from Peer Programs
Abstract:
Computer programming has become a fundamental tool that enables progress
across a broad range of disciplines including basic science,
communications, and medicine. Yet, Computer Science is failing to
attract the number of students necessary to sustain progress both within
the discipline and in those disciplines supported by computer science.
Some recent research has focused on creating programming environments
that introduce young students to computer programming in a motivating
context. One of these systems, Storytelling Alice motivates middle
school children, particularly girls, to learn programming in order to
build animated stories. In a formal study, we found that 51% of
Storytelling Alice users versus 17% of Generic Alice users snuck extra
time to keep programming. While a motivating context for learning
computer programming is necessary to increase the number of young
students who learn to program, it is not sufficient. For many pre-high
school students, formal opportunities to learn computer science simply
do not exist. We are currently working on a new system called Looking
Glass which maintains storytelling as a motivating context and focuses
on developing user interface support that enables middle school aged
children to easily and effectively teach themselves using programs
created by peers. Looking Glass will incorporate tools that enable users
to identify sections of peer written programs that interest them and
then follow automatically generated tutorials to learn how to create the
selected sections of those programs in their own context. In this talk,
I will describe our proposed framework for supporting users in learning
from peer-created programs, present results from an exploratory study of
novice programmers searching for code in unfamiliar programs and a
prototype code-finding tool.
- March 06, 2009
Speaker: Dr. Ali Butt;
Slides, and Survey Link
The "coolness" of reliability and other tales
Abstract:
As storage capacity of disks experience unprecedented growth, latent sector
errors have become the main obstacle to ensuring data integrity and
improving storage reliability. In this talk, I will discuss the latest
trends in circumventing such errors, and present the concept of energy-cost
of storage reliability. I will also present how latent errors affect modern
High-Performance Computing Centers and reduce their serviceability, and
describe a decentralized data offloading approach for mitigating the impact
of center storage failures.
- March 20, 2009
Event: Graduate Reruiting Weekend; No survey due
- March 27, 2009
Speaker: Dr. Godmar Back;
Slides, and Survey Link
Towards a new infrastructure for the World Wide Web
Abstract:
Recent years have seen a dramatic shift towards the World Wide Web as
both the deployment platform and primary user interface for many, if
not most, new applications. Yet, the existing systems infrastructure,
which has been designed for a simple hypertext system, is woefully
unprepared to support and sustain this change.
This talk will present LibX, a large-scale distributed web application
we have built, to highlight the potential benefits and uses of these
new types of applications. We will also use it to illustrate how and
why existing web browser and server technology is insufficient when it
comes to supporting these cloud applications. We discuss an ongoing
research project that creates abstractions and implementations
intended to provide sustainable solutions for these problems.
- April 03, 2009
Speaker: Dr. Barbara Ryder;
Slides, and Survey Link
Blended Program Analysis
Absract:
A new analysis paradigm, blended program analysis, enables practical,
effective analysis of large framework-intensive Java applications for
performance diagnosis. Blended analysis combines a dynamic
representation of program calling structure with a static analysis
applied to a region of that calling structure with observed
performance problems.
The initial instantiation of the paradigm addresses the issue of
performance bottlenecks stemming from overuse of temporary objects,
common in these applications. A blended escape analysis, which
approximates object effective lifetimes, has been designed and
implemented. Experiments demonstrating its utility in explaining the
usage of newly created objects in a program region have yielded
promising results (ISSTA07, FSE08). A case study on the Trade
benchmark shows how blended escape analysis helped to locate
the single call path responsible for a performance problem involving
objects created at 9 distinct sites and as far away as 6 levels of call,
in a region which calls 223 distinct methods with a maximum call depth of 20.
This talk will present the blended analysis paradigm, discuss the
blended escape analysis results obtained, new metrics designed
to measure precision and cost of the analysis, and future work.
- April 10, 2009
Speaker: Dr. Sean Arthur;
Slides, and Survey Link
The Pervasiveness of Software Engineering
Absract:
The goal of software engineering is to produce a quality product, on time,
and within budget. Over the years, many process models, activities and tools have been proposed and developed to support the software engineering effort. This presentation provides an introduction of several related research areas that focus on models and activities intended to advance the objectives of software engineering. The first research component addresses Software Quality Assessment (SQA), and the Objectives, Principles, Attributes Framework that resulted from that research effort. Observations stemming from our SQA work then leads us to the second research component, Verification & Validation (V&V), and in particular, the importance of Independent V&V. In a similar manner, results from our (I)V&V research has led us to explore the Requirements Engineering arena, and to investigate better ways to support requirements generation. This work, too, is outlined. Finally, the last two components of this presentation will focus on SE research as it relates to Agile Software Development and Software Security.
- April 17, 2009
Speaker: Dr. Cliff Shaffer;
Slides, and Survey Link
CS Educational Research at Virginia Tech
Absract: Digital education is one of the most active areas of
research within the Department of Computer Science at Virginia Tech. In
this talk, I will introduce CS education as a research topic, discussing
such issues as whether it can be considered to be Computer Science, or
even a science at all. We will see that, like all education-related
research, it is difficult to achieve meaningful pedagogical results. I will
describe a number of projects that I have been involved in, present
some results (both successful and unsuccessful), and outline plans for future
research.
- April 24, 2009
Speaker: Dr. Yong Cao;
Slides, and Survey Link
GPUs for Smarties: From Gaming to General Purpose Computing
Absract: Due to the tremendous performance improvement of Graphics
Processing Units (GPUs) in recent years, the desktop PCs equipped with
state-of-art GPUs are labeled as "Personal Supercomputer". These commoditized
GPU systems can reach a computing capacity of approximately 7 Tera FLOPs, and only cost around 4,000 dollars. However, only a certain type of algorithms, data parallel algorithms, can take full advantage of the GPU power, because of high memory latency and the lack of synchronization mechanism.
In this talk, I will discuss algorithm optimization strategies for NVIDIA GPUs, focusing on how to maximize the level of parallelism and how to hide memory latency. I will use several research projects as examples. These projects include both data-parallel applications, such as video processing and volume rendering, and non data-parallel algorithms, such as a PDE (partial differential equation) solver and temporal data mining.
- May 01, 2009
Speaker: Dr. Danfeng Yao (Rutgers);
Keystroke Dynamic Authentication With Trusted User Inputs For Botnet Detection
Absract:
Studies show that millions of computers worldwide have become bots, i.e., armies of zombie PCs controlled and used by cyber criminals to launch attacks. Most existing detection solutions closely follow how botnets behave and thus are not adaptive. We believe that there are intrinsic and fundamental differences between how a human and a bot interacts with a computer, which can be leveraged to detect infected hosts. Our approach is to monitor and analyze the characteristic human behavior patterns of the PC owner to detect anomalies.
In this talk, we describe a remote authentication framework called Telling hUmans and Bots Apart (TUBA) that extracts, analyzes, and classifies a PC owner's characteristic keystroke patterns. We further design a TUBA integrity service that uses a lightweight cryptographic verification mechanism to prevent the injection of fake input events. Our TUBA prototype is realized in a flexible client-server architecture that allows the scalable trusted monitoring. A comprehensive security analysis on the attacks and defenses of our framework is presented.
Bio:
Danfeng (Daphne) Yao is an assistant professor in the Department of Computer Science at Rutgers University, New Brunswick. She received her Computer Science Ph.D. degree from Brown University. Her research interests are in network and information security, in particular user-centric security and privacy, social- and human-behavior pattern recognition, insider threats, secure information sharing, data privacy, and applied cryptography. Danfeng has 25 publications on various topics of security and privacy. She won the Best Student Paper Award in ICICS 2006, and the Award for Technological Innovation from Brown in 2006, both for her privacy-preserving identity management work. Danfeng has one provisional patent filed for her recent bot detection techniques. She interned in the Trusted Systems Lab at HP Labs in 2005, and visited CERIAS at Purdue University as a visiting scholar in 2007. She is a member of DIMACS and DHS DyDAn Centers.
|