Research
My research experience is in HPC and scalable systems research with a focus on communication and networks. In my dissertation I examined contention in the memory subsystem as a result of one-sided communication. I studied power saving opportunities in the network fabric and how we can develop dynamic, and responsive network monitoring. In the future I would like to apply my knowledge of communications and networks to improve the performance of applications and develop efficient end-to-end workflows. More broadly, my long-term career goals are to engage in HPC research that has a direct impact application performance and scalability.
Keywords: Networks, Communications, Performance Modeling, Simulation
Contents
Academic and National Laboratories
Lawrence Berkeley Laboratory -- NERSC
Research exploring design of HPC systems with a focus on networks at NERSC
- Unraveling Network-induced Memory Contention: Deeper Insights with Machine Learning, Transactions on Parallel and Distributed Systems IEEE 2017
- Hardware MPI Message Matching: Insights into MPI Matching Behavior to Inform Design, EXAMPI in association with SuperComputing 2017, Denver Colorado
- Understanding Performance Variability on the Aries Dragonfly Network, HPCMASPA in association with IEEE Cluster 2017, Honolulu Hawaii
Sandia National Laboratories & UNM
Research in the Center for Computing Research, exploring power and performance of HPC networks.
- Doctoral Showcase Poster and Doctoral Showcase Summary, SC 2016, Salt Lake City UT
- (SAI) Stalled, Active and Idle: Characterizing Power and Performance of Large-scale Networks, IEEE Cluster 2016, Taipei Taiwan
- NiMC: Characterizing and Eliminating Network-Induced Memory Contention, IPDPS 2016, Chicago IL
- RMA-MT: A Benchmark Suite for Assessing MPI Multi-threaded RMA Performance, CCGrid 2016
- Power Aware, Dynamic Provisioning of HPC Networks, SAND2015-8717, October 2015
- BALANCING POWER AND TIME OF MPI OPERATIONS, December 2014
University of New Mexico
Research Under Dorian Arnold work focuses on high performance computing and scalability issues. In particular, I worked on development of efficient overlay networks, bootstrapping strategies, and system monitoring solutions.
- An Optimal Algorithm for Extreme Scale Job Launching, ISPA 2013
- A LogP extension for modeling Tree Aggregation Networks (TAN), HPCMASPA 2015
Texas State University
Research under Xiao Chen
Term Projects
- ROFLCOPTER: A program developed to find hidden inter-dependencies between companies on the S&P500. Term project for undergraduate data mining course. Programmed in C#.
- Comparison between Brainfuck and Completely Fair Scheduler on commodity systems - Dec 2009.
- Stupid Routing is Smart - a routing strategy for Dave Ackley's Illuminato X Machina's (Arduino-ish boards that can be joined together). To test this, a distributed prime factorization workload ran across 15 boards.
Industry
Yahoo!
- Working on the network infrastructure team at Yahoo! Inc. Work involved understanding of Yahoo's next-gen networks and deploying frameworks for collecting and storing metrics to support the operations team. Using OpenTSDB, custom switch collectors on top of a emulated cluster.
Kula Media Group
- Work on the Kulabyte Video Encoder. I spent part of my time at Texas State working with a local San Marcos company, doing QA for Windows-based video encoder development. The encoder was successfully used for live, HD, streaming of Operation Myspace from Iraq and Major League Baseball.