MCS Student Researchers Summer Presentations
TGraph: Exploring Trend Components in a Time-Varying Volumetric Data
Visualizing, tracking and understanding data evolution in a time-varying volumetric data is difficult due to visual occlusion. To address this problem, we introduce TGraph, a dynamic graph representation of the whole data set. In TGraph, a step graph is built for each time step.
It extracts features based on trends or residuals after detrending, treats the features as nodes and represents their relations with weighted edges. In addition, step graphs in adjacent time steps are connected by edges to represent data evolution. In this talk, I will explain how we construct the TGraph and introduce several graph analytic techniques that will help users better understand the data sets. At last I will cover some results and the work that we have not finished.
Yi Gu is currently a Ph.D. student in University of Notre Dame, and was a PhD student in Michigan Technological University. His research interest mainly locates in graph-based visualization.
YARNsim: Next Generation Hadoop Simulation System
Large scale distributed systems, such as Hadoop and Next Generation Hadoop: YARN, are a challenge to design and understand. These systems consist of many design points that influence application performance. While modeling and simulating these systems is useful for identifying successful system designs and application performance optimization, these activities are a significant challenge because it is difficult to accurately and efficiently model the interactions between system components. Parallel discrete-event simulation (PDES) tools provide a convenient way to accurately model complex interactions of these system components with sufficient fidelity, efficiency, and fast turnaround time.
In this talk, we present YARNsim, using Rensselaer Optimistic Simulation System (ROSS) and Co-Design of Exascale Storage System (CODES) frameworks. The goal of YARNsim is to capture the complex interactions between system components, highlight the new features introduced by YARN and keep a balance between simulation accuracy and efficiency. This talk will discuss: 1) an introduction to YARN, its major components, and the notable differences between it and previous generations of Hadoop/MapReduce, 2) a brief overview of PDES and how to build the simulation components from a real system, and 3) a description of the design and implementation plan for YARNsim using the ROSS/CODES frameworks.
Ning Liu is a second year Ph.D. student from the Computer Science Department of Illinois Institute of Technology. His research interests include parallel simulation, co-design and performance optimization of parallel and distributed systems.