James D. Teresco, Ph.D.

Interests

Data structure and algorithm visualization using highway data and mapping APIs. Parallel scientific computation. Dynamic load balancing for adaptive computations in heterogeneous, hierarchical (including hyperthreaded and multi-core), non-dedicated and transient computational environments. Distributed data structures. Tools to facilitate parallel processing. Computer Science education, particularly in relation to parallel processing.

Publications and Conferences

  • Publication List
  • Conference List
  • Software

  • Hierarchical balancing and others in the Zoltan Toolkit at Sandia National Laboratories.
  • Resource-aware balancing in the Dynamic Resource Utilization Model (DRUM).
  • zoltanParams, a parameter parsing library for Zoltan.
  • The Parallel Mesh Database (PMDB).
  • PMDBtool, a tool to help use PMDB.
  • Micro-Blogging Retrieval

    I supervised the work of MS project student Manar Alohaly in 2013. Her work was in information retrieval, in particular using popularity of publishers to estimate temporal relevance in a micro-blogging environment (Twitter). Her work was published and presented at the 52nd Annual ACM Southeast Conference in Kennesaw, Georgia, in March 2014.

    Ph.D. Dissertation

    Ph.D. Dissertation: A Hierarchical Partition Model for Parallel Adaptive Finite Element Computation, 2000, Rensselaer Polytechnic Institute. Advisor: Joseph E. Flaherty. [Details]

    My Ph.D. work was done in the Department of Computer Science at Rensselaer Polytechnic Institute, working with Dr. Joseph Flaherty, whose group developed software to solve systems of partial differential equations on parallel machines using adaptive finite element and related methods. I was involved in several projects, including a generic 1-D PDE solver, looking into partition quality metrics for mesh partitions, and helping to maintain and enhance software for partitioning and dynamic load balancing of three-dimensional meshes in a parallel adaptive environment. One of our favorite test problems involved simulating the flow in the perforated muzzle brake of a cannon. My thesis work included the design and implementation of the Rensselaer Partition Model, which includes the parallel mesh infrastructure used by several projects at RPI's Scientific Computation Research Center.

    Sandia Collaboration

    I had a multi-year collaboration with Sandia National Laboratories in Albuquerque, New Mexico, and spent most of my leave during the 2003-04 academic year as a Computer Science Research Institute visitor.

    Hardware

    I managed the Bullpen Cluster at Williams College for several years. The cluster was built with Sun hardware. Some of the Bullpen nodes were later integrated into the Dhanni Cluster at Williams, and smaller teaching clusters I constructed at Mount Holyoke and at Siena.