2016 Summer Student Presentations

Event Sponsor: 
Argonne Leadership Computing Facility Summer Student Presentations
Start Date: 
Aug 17 2016 - 9:30am
Building/Room: 
Building 240/Room 1404-1405
Location: 
Argonne National Laboratory
Speaker(s): 
Fabio Miranda / New York University
Danil Kaliakin, University of Nevada – Reno
Rahul Dendukuri/ Northern Illinois University
Clay Kauzlaric / Northwestern University
Yuping Fan / Illinois Institute of Technology
Takanori Fujiwara / University of California – Davis
Bharat Kale / Northern Illinois University
Bhavya Kumaran / Case Western Reserve University
Junteng Jia / Cornell University
Edouard Brooks / The University of Chicago
Sophia Chun / The University of Chicago

Welcome & Opening Remarks
9:30 am – 9:35 am Dr. Michael E. Papka / LCF Division Director

Time: 9:40 am – 9:50 am
Presenter: Fabio Miranda
Education: PhD 2018, Computer Science, New York University
Mentor: Venkat Vishwanath

Hypothetical Cities: Changing City Morphology to Minimize Adverse Weather Effects
Cities have been the center of economic activity and human development throughout the ages. By 2050, it is
expected that up to 90% of the population in North America will reside in urban areas. With an ever
increasing urbanization process, it is important that different stakeholders (city officials, urban planners,
climate scientists, etc.) take into consideration the impact that urbanization and growth has on the local
weather of a city. Phenomenons such as heat island are directly connected to how we choose to use the land
within the city limits. Recent technological innovations have enabled advanced weather simulations that take
into account the shape of the city (e.g. building heights) and how it is currently being used (e.g. building
energy output), allowing a detailed study of how the urbanization process is impacting local weather. We
leverage such innovations, coupled with visualization techniques, to propose a framework that allows for the
exploration of hypothetical cities. Our goal is to understand how a change in the urban morphology of
existing cities can positively (or negatively) impact the weather, and, hopefully, guide future urban
developments in such a way that adverse phenomenons are minimized.

Time: 9:55 am – 10:05 am
Presenter: Danil Kaliakin
Education: PhD 2018, Chemistry, University of Nevada – Reno
Mentor: Yuri Alekseev

Implementation of Fragment Molecular Orbital Method to a Minimum Energy Crossing Point Search in
Quantum-Chemical Package GAMESS

Many of the important industrial processes require usage of the rare noble metals for catalysis. In contrast,
metal-sulfur proteins (MSPs) are utilizing abundant first-row transition metals to catalyze chemical reactions.
Understanding how MSPs work is a major step toward development of new generation of inexpensive and
efficient industrial catalysts. The crucial parameter in the description of electronic transitions in MSPs is the
minimum energy crossing point (MECP). The goal of this project is to couple the MECP search algorithm with
the highly parallel fragment molecular orbital (FMO) method for studying of spin-forbidden processes in
MSPs by using parallel computers

Time: 10:10 am – 10:20 am
Presenter: Rahul Dendukuri
Education: MS 2016, Computer Science, Northern Illinois University
Mentors: Mark Fahey & Mike Papka

Part-Replacement/Failure Analysis and Prediction for Mira & improvements for the “Gronkulator”
With the expiration of operational support from IBM for the BlueGene/Q supercomputing systems, the
availability of sufficient spare parts has become an important factor. I will present my statistical model
that is based primarily on the node-replacement history of the machine since the beginning of
production. The node failure/replacement trends have been visualized and the lifetime of the machine
can now be estimated through the proposed linear regression model.
The second part of my talk will be about my effort to improve the current version of the ALCF Machine
status information page known as “The Gronkulator” (www.status.alcf.anl.gov) to better suit the
architectural and operational requirements of the Cooley cluster as well the computing resources to be
installed in the future –– starting off with Cray XC40 (Theta).

Time: 10:25 am – 10:35 am
Student: Clay Kauzlaric
Education: BS 2018, Computer Science, Northwestern University
Mentors: Mike Papka & Rajesh Sankaran

Waggle: An Open Platform for Intelligent Attentive Sensors

Waggle is a research project at Argonne National Laboratory with the aim to create a networked, extensible,
low power in-situ computing capable sensor platform ideal for environmental sensing as well as urban
research in smart cities. The project’s hardware relies on efficient and powerful single-board computers, a
custom designed power management board, and leading edge sensor technologies. These powerful physical
components are paired with the plentiful resources of cloud computing and open-source, modular software
to create an ideal stage for researchers to customize their sensor network by adding their own sensors and
data analysis methods. The Waggle platform is currently implemented in Array of Things, an urban sensing
project for the city of Chicago, with plans to deploy in the Chicago Botanic Garden, Cook County Forest
Preserve, and the Indian Boundary Prairies. In this talk, I will present my involvement in the project, including
assembly, testing, debugging, procurement, and deployment.

Time: 10:40 am – 10:50 am
Presenter: Yuping Fan
Education: PhD 2020, Computer Science, Illinois Institute of Technology
Mentors: Bill Allcock & Paul Rich

Adding Data Staging Functionality to the Cobalt Scheduler

The big picture of this project is to build a general framework that can make integration of future
“dimensions” easier and more consistent. For the summer project, the data staging functionality will be
added so that there will be the command line option that allow user to transfer files and then run jobs
automatically. Specifically, functionality could be added to process arguments, hold user job, stage data in
and monitor the progress of the transfer. Optionally stage results on when the job completed.

Time: 11:00 am – 11:10 am
Presenter: Takanori Fujiwara
Education: PhD 2020, Computer Science, University of California – Davis
Mentor: Preeti Malakar

A Visual Analytic System for Massively Parallel Communications on the Blue Gene/Q Supercomputer

Faster communications among MPI processes on several thousand compute nodes is crucial for achieving
high efficiency of parallel applications. However, it is difficult to identify the communication bottlenecks due
to the large number of nodes and complex network topologies of the supercomputers. In this work, we
propose a visual analytic system that aids in identifying the communication bottlenecks in massively parallel
applications/benchmarks on the Mira Blue Gene/Q supercomputer. We also provide the ability to
interactively analyze the source of bottlenecks and enable the user to identify alternate communication
options.

Time: 11:15 am – 11:25 am
Presenter: Bharat Kale
Education: MS 2017, Computer Science, Northern Illinois University
Mentors: Janet Knowles & Mike Papa

Graph Plotting API for Apple TV Apps

The goal is to design an independent plotting library that can be used to create different visualizations of
given data through graphs. The primary purpose of this API is to create an analytical tool that can be used for
various business intelligence scenarios. There are few such libraries available, but none of them are targeted
for tvOS. We have used Swift Programming language for the development and targeted both iOS and tvOS
platforms. This library currently supports bar graphs (horizontal and vertical), line graphs and scatter plot and
can be used for displaying them on the devices using Apple TV. The API can also be easily extended to include
new graphs and other functionality and also target other platforms from Apple.

Time: 11:30 am – 11:40 am
Presenter: Bhavya Kumaran
Education: BS 2019, Computer Science, Case Western Reserve University
Mentor: Beth Cerny

Media Projects at the ALCF

During the first half of the program, I focused on creating ePosters for research projects that have allocations
at ALCF. Later, I focused on learning how to create a website using WordPress and coding in HTML. The
website created was the new ‘Exascale Age’ website for the upcoming Exascale reviews. In the second half of
the program, I am working with the motion detectors in the Visual Lab and modifying code to fit the
parameters specifically for Argonne’s display. The goal is to use devices such as Wii Remotes to control the
images on the big display wall.

Time: 11:45 am – 11:55 am
Presenter: Junteng Jia
Education: PhD 2010, Chemistry, Cornell University
Mentor: Alvaro Vazquez-Mayagoitia

High Performance Preconditioner for Solving Elliptical Poisson's Equation (Ax=b)

Solving Poisson's equations is one of the most intense computations in periodic Hartree-Fock calculations and
is ubiquitous in scientific and engineering simulations. Preconditioning techniques are widely used with
sparse linear solvers to improve the condition number of a sparse matrix. In this presentation, I will review
the existing preconditioner for solving a symmetric and positive definite (SPD) matrix generated from a
Poisson's equation. Then I will introduce a new preconditioner based on the symmetry of the partial
differential equation (PDE) matrix. The performance of my new preconditioner is compared with existing
preconditioners both in terms of effectiveness (number of iterations) and cost (time to solution), by using
efficiently vectorization and multiple threads in many-core hardware.

Time: 12:00 pm – 12:10 pm
Presenter: Edouard Brooks
Education: BS 2016, Computer Science, The University of Chicago
Mentor: Joseph Insley
Introductory A/VR projects for Scientific Visualization

With the onset of consumer augmented and virtual reality (A/VR), a growing interest has been placed in the
visualization of data on head-mounted displays (HMDs) such as the Google Cardboard, Oculus Rift, HTC Vive,
and Microsoft Hololens. Following, several introductory projects are being created to comprehend the
potential of each of these displays for SciVis, and benchmarking projects for the Cardboard and Hololens (of
which, the later contains the proprietary Custom-built Microsoft Holographic Processing Unit). Additionally, I
have been working in collaboration with LDRD researchers to present existing data in a VR environment,
currently focusing on Exif data extraction for 2d photo alignment in 3d space.

Time: 12:15 pm – 12:25 pm
Presenter: Sophia Chun
Education: BA 2018, English Literature, The University of Chicago
Mentor: Jim Collins

The Language of Science

I take information from various research papers, quarterly reports, and grant proposals and turn them into
short articles. These articles help potential investors, funding agencies, elected officials, and the general
public to understand the potential real-world impact of highly technical projects. In my presentation, I will go
over how an article is put together, such as how pieces of information are selected, and why wording as well
as conciseness is important in presenting technical information to a non-technical audience. I will emphasize
the importance of clarity and advise on how others can learn to effectively “sell" their science.