Optimizing Linux for 16-socket servers
Assisted in determining the cause of major network performance issues involving large socket count servers with greater than eight NUMA nodes and Mellanox based 100GB Ethernet controllers. The process involved optimization of the Linux kernel in the areas of interrupt handlers. The current Linux kernel can only automatically optimize systems with up to eight NUMA nodes, requiring the manual placement of interrupt handlers to extend across a 16-socket system. There are also limitations in the kernel that permit only 64 total interrupt handlers per PCI device, causing major issues with 32-socket systems.
Developing a Quantum Simulator and Software Stack
Assisted in the development of the Atos Quantum Learning Machine, a hardware agnostic quantum programming language and simulator appliance. The tasks related to the project were the optimization of the command line interface and detection of inconsistencies between different simulators and modules. The Quantum Simulator Software Stack is now available as an OpenSource project from Atos. You can find more information on MyQLM at https://atos.net/en/lp/myqlm.
A Study on Public Cloud for HPC Workloads
Completed a two year study on the performance and cost metrics of utilizing public cloud for execution of MPI based High Performance Computing Workloads.
Introduction to Parallel Programming
Parallel Programming with MPI simplified for K-12 education level. Please check out my books link to pickup a copy.
Grid Based Optimization for Next Generation Space Exploration (GridOptZ)
Project utilizing the Parabon Grid Engine to simulate various models of the Golubovic Rotating Space Elevator. This project consists of three documents the finalreport.pdf, detailedreports.doc, and the RSEcompendium.doc.
GPU LIDAR Processing
A proposal was written and rejected in an attempt to gain funding for future research into the field of LIDAR data processing utilizing the Graphics Processing Unit. I developed a working code base that can preform LIDAR data feature extraction to locate buildings, roads, and trees as well as perform bare earth data extraction in a fraction of the time required using conventional processing methods. The code base I currently have will allow for a complete bare earth extraction from a 40 billion point dataset in less than a minute utilizing the GPU versus 30 minutes utilizing conventional methods.
Cluster 2005 Conference Submission
This paper was presented to the Cluster 2005 Conference. The paper was rejected as the conference felt it was geared more toward a Visualization Conference. The lead personnel on the paper left the employment of the organization prior to resubmitting the paper for the Visualization Conferences.
This paper was written to obtain funding for the LIDAR Data Research Project. The Cluster 2005 paper and the WV View paper are a result of this work. See http://sbir.gsfc.nasa.gov/SBIR/successes/ss/5-122text.html for more information on this project. Work was done in conjunction with WVHTC Foundation, Galaxy Global Corp and Prologic, Inc.
OpenGIS Movie Implementation Specification
This specification extension was submitted for approval to the Open Geospatial Community in March of 2006. It is in a three week review process for final publication. My employer dropped interest in perusing this publication and the specification extension was never presented to the OpenGIS conference. It has been trapped in the review process for over a year. See http://www.opengis.org
Draft WMS Implementation Document
This is a preliminary document detailing the research of the WMS protocol prior to the recommendation for the Proposed Extension above. The document details the lessons learned as I pursued creating my own WMS server from various GIS software packages.
Rolls for ROCKS Clusters
I am active in the roll development community for Rocks Cluster as one of the founding members of the Rocks Community Rolls Working Group http://rocks.stanford.edu