Research Interests
Architecture independent parallel algorithm design and
implementation,
high productivity/performance cluster computing,
models of parallel computation,
Bulk-synchronous parallel computing,
interprocessor communication network performance assessment,
experimental algorithmics, graph theory and combinatorics.
Current Projects
- PowerGrid Project Cluster-based implementations of power-flow
analysis and contingency analysis algorithms and benchmarking for the
development of FPGA-based bulk-synchronous parallel algorithms. This is
joint work with Prof. S. Ziavras at the ECE Department at NJIT (NJIT group)
and Drexel University.Research supported by a Department
of Energy (DOE) grant
- High Performance and Productivity Computing Systems for space weather
research. This is joint work with colleagues in the Physics Department
at NJIT (the space weather part) and the CS Department (the image processing
part). Research is supported by NSF ITR grant (Sep 1, 2003- Aug 31,
2007) IIS-0324816 in which i serve as a Co-Principal Investigator .
- Parallel Algorithms in Finance Design and implement parallel
algorithms for finance-based problems such as option price valuations
on binomial and trinomial trees either directly or using finite difference
methods (explicit and implicit). Design algorithms in an architecture
independent way and suited for high-latency systems such as PC clusters.
This research was partially supported by NSF/MRI grant NSF EIA-9977508
that run from Sep 1, 1999 through Aug 31, 2003, in which i served as
a Co-Principal Investigator.
- Latency Tolerant algorithms for matrix computations
Design and implement latency tolerant algorithms for dense and sparse
matrix computations and FFT problems.
This was also partially supported by NSF/MRI grant NSF EIA-9977508.
- Internal and External Memory Sorting
Design, analyze and examine parallel performance of deterministic
and randomized sorting algorithms on cluster of SMP workstations.
- Parallel Algorithms for Proximity Searching in large texts
and GIS systems.
Joint work with M. Marin (U. de Magallanes, Chile).
Research Results
In the past three years I have been involved in
architecture independent design, analysis and implementation
of parallel algorithms on a cluster of PC workstations that has
been set up at the CS Department at NJIT under the support
of NFS under Grant MRI NSF EIA-9977508 and SBR grant 421350.
Information about the PC cluster lab, research results obtained,
and publically available source code can be obtained through the links
below.
-
PC Cluster Laboratory (CS Department) The laboratory was
established with support from NSF MRI EIA-9977508 Grant and matching
funds from NJIT, as well as support from grant SBR 421350.
This link is under construction and updated frequently.
-
General Software Code
that is being used in published or submitted papers.
-
Parallel Software Code
Building architecture independent and
parallel library independent parallel code.
The introduced algorithms are designed and analyzed in an
architecture independent way under the Bulk-Synchronous Parallel
(BSP) model of computation. In order to support and verify
the theoretical results we implement many of the parallel algorithms
using various communication libraries.
The parallel codes we have been developing can be run under
LAMMPI and BSPlib. Only recompilation of the same source is required
for our code to run on one or the other library. This
communication library portability comes at no significant cost
as attested by the comparable and high performance of the implemented
algorithms.
-
Recent Publications
Last Update : May 7, 2010