1. CS @ ILLINOIS Visualizing Graphs Distributed Across Multiple Processes
Morgan Wang & Kartikey Desai
Mentor: Amanda Bienz
Department of Computer Science
Visualizing Graphs Distributed Across Multiple Processes
BACKGROUND
Graph partitioning is particularly important to parallel programming, since most mathematical problems, and therefore scientific problems, can be thought of in terms of graphs
If a graph is partitioned well (meaning equal distribution of nodes and few edges across processors), the program will be MUCH more efficient than a poorly-partitioned graph
Our projects aim to visualize these factors to help determine whether poor performance is due to graph partitioning
KARTIKEY’S PLOTTER
The plotter displays the nodes and edges involved in a single process
Nodes represent individual values/problems
Edges were color-coded to help visualize how many edges are present within a single process, and between multiple processes
MORGAN’S GUI
I worked on the overall statistics of the partitioned graph
Each circle represents each process and it’s size depends on how many nodes that process is processing
Each edge represents connections between processes
WHAT WE’VE LEARNED
How graph partitioning is utilized in parallel processing problems
How graphs are used to represent complex math and science problems
Proper research practices, and how to collaborate on code in a team
Gained experience in: HTML, CSS, JavaScript
Figure 2
WHAT COMES NEXT
Develop a more uniform mechanism for feeding data to the graphs (i.e. a matrix parser that reads in from a data file)
Link the two programs so when one node is clicked, the plot of its nodes can be shown
Giving GUI the ability to modify the input data in a visual manner