SciClone Cluster Project Computational Science Cluster
Home
Introduction
Sponsors
Research
Hardware
Software
User Info
Documentation

Research Projects

A partial listing of research projects conducted with the SciClone cluster, organized by topic:

Previous projects are listed here.

Applications

Simulations of NMR Relaxation and Lineshapes (R. L. Vold, 5/02)
Matching experimental NMR lineshapes to those simulated using either the Liouville-Von Neumann equation (no motion) or the stochastic Liouville equation (random motion) provides quantitative characterization of molecular motion in complex, technologically useful materials. The goal of this project is to improve and expand the scope of simulation procedures by taking advantage of parallel processing. This will be especially useful for powder samples, since the calculations for each randomly oriented crystallite in the powder are independent of any other and can therefore be done concurrently.

Study of Multiquark Systems within a Constituent Quark Model (V. Nazaryan, 10/03)
In light of the recent discovery of a new particle made of five elementary constituents (5 quarks), the understanding of symmetries and various other properties associated with multiquark systems has gained renewed interest. Our work focuses in particular on the study of such systems within the framework of a constituent quark model. This involves solving complicated group theoretical problems when constructing wave functions of multiquark systems that have to obey the fundamental symmetries of nature. The use of computational methods greatly simplifies and accelerates the process of solving these and related problems.

Ultra-Short Laser Pulse Beam Shaping (S. Zhang, 2/04)
We are studying the spatio-temporal intensity profile of an ultrashort laser pulse passing through a laser beam shaping system, which uses diffractive optical elements to reshape the Gaussian profile of a laser beam into a flat-top distribution. The Nonlinear Schrödinger Equation is solved numerically to simulate the nonlinear optical effects in this system. These computational results will be compared with experimental data for 136-femtosecond pulses at 800 nanometer wavelength from a Ti:sapphire laser.

[200401-01] The Ground State of Trapped Bosons with Attractive and Repulsive Interactions (W. Purwanto, 1/04)
We study the ground state of trapped bosons with attractive and repulsive interactions using the Auxiliary-Field Quantum Monte Carlo method.

Kinetic Monte Carlo Simulations of Crystal Growth in Ferroelectric Materials (M. Suewattana , 7/02)
We study the growth process of single crystal perovskite alloys by Kinetic Monte Carlo (KMC) methods. An ionic model with long-range Coulomb interactions is used to model these relaxor single crystals. An enhanced KMC algorithm has been developed to handle the long-range interaction. We initially focus on the growth of heterovalent binary crystals.

[200308-02] Quantum Monte Carlo Study of Strongly Correlated Systems (W. Al-Saidi, H. Kwee, 8/03)
This project applies density functional theory and auxiliary field quantum Monte Carlo methods for studying systems of atoms and molecules as well as bulk properties of solids. The results, besides being interesting in their own right, aim at a better understanding of the correlation energy in strongly correlated systems, as well as shedding light on some of the fundamental properties of several systems.

First-principles Calculation of Piezoelectricity in Monoclinic PZT and PMN (Z. Wu, 8/01)
First-principles calculation, LAPW and Mixed Basis methods, and the modern theory of piezoelectricity are applied to the newly found monoclinic phase of PbZr1/2Ti1/2 and PbMn1/3Nb2/3. We will compute the intrinsic and homogeneous piezoelectricity response from ground-state Berry's phase calculation. We want to elucidate the gigantic piezoelectric constants of PZN-PT and PMN-PT from first principles.

Lattice Boltzmann Modeling for Magnetohydrodynamics (A. Macnab, 2/01)
A variety of physical systems can be modeled with highly parallelizable computational schemes known as Lattice Boltzmann Methods (LBM). We are investigating the use of LBM's for magnetohydrodynamic systems. The present model uses an octagonal lattice geometry that is believed to eliminate many of the non-linear instabilities present in earlier models while maintaining 75% of the computational efficiency.

Autoresonant Excitation of Nonlinear Waves (C. Kulp, 7/02)
In this project we are attempting to develop a simple controller for nonlinear wave systems, using the Nonlinear Schrodinger Equation (NLS) as the model system. The NLS is the generic equation for waves which to leading order are weakly nonlinear and monochromatic. We've had success controlling the NLS to one of its known solutions q0. The solution of q0 is a plane wave near t=0 and then as time progresses an instability causes q0 to form a pulse, saturate, then return to a plane wave. We have also been able to control other related nonlinear wave systems (Ginzburg-Landau equation and Dysthe's equation) to this solution. We will try to apply this control scheme to additional nonlinear wave systems such as Zakharov's equations for plasmas. We are also interested in developing simpler controllers which yield the same results.

The Opium Project (E. Walter, 9/02)
Opium is an open-source pseudopotential generation package with various developers at William and Mary and the University of Pennsylvania. One of the main goals of the project is to provide compatibility across the most popular computer platforms. SciClone will be used for development and testing of the Sun/Solaris version.

Epistrophic Maps (S. Knudson, J. Delos, 6/01)
We are looking for order within chaotic structures as generated by mapping processes which are designed to resemble the analytic results of solving Newton's equations of motion. Such maps have not been thoroughly explored, and all aspects of the problem must be resolved. The numerical problem involves both the generation of large datasets and visualization of the results.

[200312-01] Three-Dimensional Ocean Model of Hydrodynamics and Sediment Transport (C. Harris, A. Bever, C. Sherwood, J. Rinehimer, M. A. Bynum, 12/03)
Predicting sediment transport in coastal oceans requires that we account for processes that erode, supply, and redistribute sediment, including storms, floods, winds, tides, and energetic waves. One- and two-dimensional models that work well in other types of coastal areas poorly represent areas offshore of river mouths, which are inherently three-dimensional, and also fail in areas of strong coastal currents. Three-dimensional hydrodynamic models provide a promising platform within which to predict sediment suspension and transport in such areas because they already account for ocean currents and plume dynamics. This work proposes to improve an existing three-dimensional sediment transport model, initially using data gathered in the Adriatic Sea. Simulations to date have been limited by computational speed and specification of forcing waves. Seasonal time-scale simulations could be completed on available supercomputers, and we plan to attempt this on the SciClone system. Specifically, we will be using ROMS (Regional Ocean Modeling System) and NCOM (Navy Coastal Ocean Model) .

Evolutionary Dynamics of Enzyme Molecules (C. Bagdassarian, D. Kranbuehl, S. Shang, A. Hossain, 2/01)
We will be simulating the evolutionary dynamics of a fluctuating "toy" enzyme. Molecular Dynamics runs for the fluctuation dynamics will be coupled with genetic algorithms for the mutation and evolution of the enzyme. We hope to show that strongly correlated fluctuation patterns arise in maximally efficient catalytic enzymes.

Benchmarking NAMD: Simulations of Biomolecular Systems (J. Phillips, G. Zheng, 2/01)
NAMD is a parallel, object-oriented molecular dynamics code designed for high-performance simulation of large biomolecular systems. NAMD is distributed free of charge and includes source code. We will benchmark the performance of NAMD on SciClone for comparison to other platforms.

[200307-02] Use of MALDI-TOF Mass Spectra for Disease Diagnosis (E. Tracy, D. Cannan, 7/03)
This is a pattern recognition project in collaboration with researchers at INCOGEN and EVMS. We are manipulating large data files generated by mass spectrometers to look for biomarkers for disease.

[200303-01] Mass Spectrometry Data Analysis for Proteomics (D. Malyarenko, D. Manos, W. Cooke, H. Chen, K. Miller, L. Wei, R. Marchitelli, 3/03)
Design and testing of signal processing and classification modules in the shared software framework, created by INCOGEN, for EVMS mass spectrometry data analysis and cancer diagnosis.

Simulation of Ion Plumes Generated by Laser Ablation in SELDI/MALDI Mass Spectrometers (E. Tracy, 7/03)
This is a theoretical modeling project in support of experiments to be performed later this summer in the William and Mary Physics Department, in collaboration with INCOGEN. The goal is to develop an improved mass spectrometer for disease diagnosis and proteomics research.

Electrodiffusion of Calcium Near a Point Source (R. Ellison, 1/03)
An analytical study of buffered diffusion and electrodiffusion of calcium in living cells.

[200107-01] The Effect of Feedback Inhibition on Sensory Relay by the Visual Thalamus (G. Smith, S. Coombes, K. Srinivasan, P. Brewer, M. Huertas, J. Groff, 7/01)
This project involves the development of a network model of the relay of retinal information to the cortex by the dorsal lateral geniculate nucleus (dLGN) that accounts for the feedback from inhibitory neurons located in the perigeniculate nucleus (PGN). This model is being used to test the hypothesis that the state of these inhibitory neurons (reflected in burst vs. tonic responses to excitation) influences the dynamic filter properties of the dLGN/PGN "relay".

[200206-01] The Dynamcs of IP3-Sensitive Ca2+ Release Sites (G. Smith, V. Nguyen, C. Loomis, C. Wei, B. Mazzag, J. Zhang, 6/02)
Computational models of the dynamics of Ca2+ regulation at IP3-sensitive Ca2+ release sites are being developed focusing on the stochastic activation and inactivation of IP3Rs consistent with current knowledge of IP3 diversity and a realistic account of the buffered diffusion of intracellular Ca2+ leading to cooperative IP3R activity.

Emergent Properties in the Respiratory Rhythm-Generating Neural Network of Mammals (J. Hayes, 6/04)
We are simulating networks from the mammalian brain stem that are responsible for the generation and control of breathing rhythms. We are particularly interested in how synaptic connection patterns affect the ability of the networks to produce synchronized rhythmic activity when all neurons have generic excitability properties and excitatory-only synapses.

A Novel Method for Bolstering Software Reliability (A. Mundra, 4/04)
Traditional testing can not prove the validity of critical software whose malfunction may have severe consequences. On the other hand, exhaustive testing is infeasible. One way to bolster testing is to employ assertions as runtime checks, thereby relieving the tester of the need to determine the correct output for every input to the system—the assertions provide a self-check. We propose a new approach to testing in which assertions are written to provide a self-check, and then the system is executed for a large number of inputs. We will evaluate our approach by applying it to a case study. SciClone will be used mainly to generate inputs and conduct bounded exhaustive testing.

Symbolic State Space Exploration in S.M.A.R.T. (R. Marmorstein, R. Siminiceanu, 2/02)
Model checking and state-space exploration applications suffer from the state-space explosion problem. Our tool, SMART, uses symbolic techniques to tackle this problem. In order to explore suitably complex systems, we intend to exploit the resources of the SciClone cluster.

Parallel Algorithms and Numerical Methods

Development of Parallel Genetic Algorithms for Recurrent Neural Networks (S. Stallman, 12/04)
We are investigating genetic algorithms to evolve a Recurrent Neural Network topology that has both long term and short term learning abilities. The search space is evolved in parallel, with parallel optimization of the Genetic Algorithm as a target concern..

Decomposition Methods for the Optimization of Coupled Systems (M. Lewis, 9/01)
The optimization of complex, coupled systems arises frequently in engineering design. We are investigating whether it is possible to devise efficient, provably convergent nonlinear programming approaches that take advantage of the decomposition of such problems along the lines of the constituent pieces.

Asynchronous Parallel/Distributed Pattern Search Algorithms (P. Shepherd, 5/02)
The goal of our research is to develop a new class of adaptive asynchronous parallel and distributed pattern search algorithms for constrained optimization problems. The target environment will be heterogeneous clusters (e.g., computational grids). These new algorithms will adapt dynamically to changes in computational resources.

Comparison of Zero-Order Optimization Methods (A. Yates, 6/02)
Several software packages which use zero-order methods to optimize user functions will be compared. Specifically, we are interested in seeing how each package performs when a solution is on or near the defined boundaries. SciClone will be used to collect accurate timings as one measure of software effectiveness.

[200307-01] A Robust and Efficient Solver for Symmetric Eigenvalue Problems (J. McCombs, 7/03)
This project focusses on developing a robust Jacobi-Davidson algorithm for solving difficult eigenproblems. The research is divided into three parts: development of a multigrain parallel algorithm for reducing the effects of network latency; improvement of restarting techniques, locking algorithms, and convergence criteria; and derivation of defaults that promote ease of use while maintaining competitive performance.

[200203-01] Dynamic Load Balancing in a Coarse-Grain Parallel Iterative Eigenproblem Solver (R. Mills, H. Jiang, 3/02)
In computing environments with SMP's or timeshared clusters of workstations, there may be multiple jobs running on a given node and/or the cluster may be heterogeneous, so load balancing becomes a very important issue. We have adapted a recent coarse-grain parallel implementation of a Jacobi-Davidson eigenproblem solver to utilize a dynamic load balancing scheme that is effective in timeshared and heterogeneous environments.

[200308-01] Dynamic Application-Level Adaptation to CPU and Memory Load in Scientific Applications (R. Mills, 8/03)
The performance of parallel scientific codes on non-dedicated computational resources may suffer greatly due to contention with other jobs for limited memory and CPU resources. It is our thesis that the dynamic, unpredictable nature of such computing environments necessitates a degree of system adaptivity within computational applications. We pursue this thesis on two complementary fronts: 1) load balancing within iterative methods via algorithmic modifications, and 2) development of a portable framework for graceful adaptation to memory pressure within scientific and data-intensive applications.

[200307-03] Guaranteed Quality Parallel 3D Delaunay Mesh Generation (D. Nave, N. Chrisochoides, 7/03)
The solution of partial differential equations (PDEs) by the finite element method requires the solution domain to be decomposed, or meshed, into geometrically and topologically simpler elements like triangles and tetrahedra. Although parallel field solvers exist, the efficiency of the overall simulation environment is often reduced by having to sequentially mesh (and remesh) the domain. In this project, we are continuing the development of GQ-PDMG, or guaranteed-quality parallel Delaunay mesh generator. GQ-PDMG utilizes properties of the Delaunay triangulation boh to efficieintly generate a distributed mesh, and to guarantee the quality of the elements in the resulting mesh.

[200309-01] Parallel Mesh Generation for Distributed Memory Machines (C. Verma, 9/03)
Mesh generation is one of the fundamental and most critical components in the simulation of natural phenomena using finite-element or finite-volume schemes. Delaunay methods are perhaps the most popular for generating 2D/3D meshes in arbitrary domains, due to their sound mathematical properties. For realistic simulations, scientists and engineers demand a very fine mesh in the domain of interest, which is very expensive to compute on sequential machines. Implementing mesh generation on shared memory machines is relatively easy, but it is a non-trivial task on distributed memory machines. We are, therefore, developing software infrastructure to support scalable mesh generation on distributed memory systems.

SciClone's Big "Brain" for "Rocket" Science (Y. Ito, A. Erukala, N. Chrisochoides, 3/04)
In this project, researchers at the University of Alabama at Birmingham and the College of William and Mary are collaborating to generate very large meshes using the Parallel Advance Front Technique. Datasets include (1) a human brain which comes from actual MRI data provided by Cornell Medical School, and (2) an exemplar of a regenerative cooled pipe geometry which comes from rocket engine testing at NASA's Marshall Space Flight Center.

[200403-01] Load Balancing of Parallel Mesh Generation for Applications from Health Care and Aerospace Engneering (G. Zagaris, 3/04)
In this project, we will be working on dynamic load balancing of parallel mesh generation for (1) a human brain which comes from actual MRI data provided by Cornell Medical School, and (2) an exemplar of a regenerative cooled pipe geometry which comes from rocket engine testing at NASA's Marshall Space Flight Center.

[200108-01] Parallel Meshing: An Evaluation (A. Chernikov, C. Verma, 8/01)
In this project we will study the performance of parallel unstructured meshing methods with respect to two fundamental issues, concurrency and synchronization. We will also develop a taxonomy in order to clarify the similarities and differences and point out the strengths and weaknesses of various classes of parallel meshing methods.

[200402-01] Longest-Edge Bisection Algorithms for the Parallel Refinement of Tetrahedral Meshes (D. Pizarro, C. Calderon, 2/04)
We are developing new parallel algorithms for mesh generation in 3D, using the concept of Longest Edge Propagation Path (LEPP) introduced by M. C. Rivara. This is joint research between the College of William and Mary and the Universidad de Chile.

[200304-01] Parallel State Space Generation (M.-Y. Chung, 4/03)
The purpose of this project is to explore techniques for generating large state spaces in a distributed fashion. The general idea is to partition Multi-value Decision Diagrams (MDD's) horizontally, then assign one or more MDD levels to each processor. Two different processors' consecutive MDD levels are connected by unsure indices, using MPI for interprocessor communication. Growth of non-canonical MDD levels can be monitored by forwarding arcs, with cleanup being done independently by level.

Parallel State Space Generation in Shared Memory Environments (J. Yu, 4/03)
In this project, we are trying to efficiently parallelize state space generation in shared memory environments. Our tests will be run on two-processor and four-processor nodes of the SciClone Cluster. We'll measure the speedups that we can get from the parallelization. The project may lead to further research in parallel model checking.

Evaluation of the Stability of ETAQA (E. Hewett, 2/03)
ETAQA is an aggregate-based approach that is used to solve M/G/1-type processes and has been extended to solve general QBDs as well. ETAQA has been shown by experimental analysis to be numerically stable and very efficient for solving both small and large M/G/1 or QBD processes. Comparisons of the ETAQA solution with Ramaswami's recursive formula will be made in order to confirm the stability of ETAQA.

ATLAS: Automatically Tuned Linear Algebra Software (C. Whaley, 6/03)
The ATLAS project is an ongoing research effort focusing on applying empirical techniques in order to provide portable performance. At present, it provides C and Fortran 77 interfaces to a portably efficient BLAS implementation, as well as a few routines from LAPACK. We are using SciClone to help get ATLAS better tuned for the UltraSPARC III processor.

Tools and Techniques for Parallel and Distributed Environments

[200302-01] VIBE Web Services (M. Levitt, Y. Li, D. Fuquay, J. Miller, K. Miller, 2/03)
VIBE is a next-generation visualization and genomics discovery tool developed by INCOGEN in collaboration with TimeLogic Corporation. Under a grant from the National Institute for Standards and Technology, INCOGEN is developing an open communications protocol for life science applications based on XML and web services. SciClone will be used as a testbed to demonstrate/evaluate this technology through the VIBE interface, serving as a web portal for internal and external users and providing computing cycles for genomic analyses.

I/O Performance Analysis of Parallel BLAST (J. Glover, 4/03)
Goals of this project include providing a parallelized BLAST algorithm for use on SciClone and gaining insight into the performance-related aspects of the system and application. We will be making use of mpiBLAST, which originated at Los Alamos National Lab.

Coordinating Distributed Bioinformatics Resources (K. Miller, 12/01)
Virginia's Commonwealth Technology Research Fund has awarded a joint grant to William and Mary, the Virginia Bioinformatics Institute, and INCOGEN Inc. to create a cooperative bioinformatics program. As part of this program, software developers at INCOGEN will utilize SciClone as a research platform for accessing and coordinating bioinformatics algorithms and resources via a web-services paradigm.

dSOARS: A Framework for Optimizing Distributed Application Performance (P. Vaidyanathan, E. Smirni, 4/03)
Workload characteristics for bioinformatics applications are different from other existing scientific applications running on distributed environments. Our framework is designed to optimize the performance of data-intensive distributed applications using replication and movement of data and code to the most appropriate location.

Willow: Survivability for Large-Scale Distributed Applications (P. Varner, 11/03)
Willow is a system designed to allow for scalable management of large-scale distributed applications for critical infrastructure. The core of the system is a decoupled communications layer that allows targeted, scalable communication based on arbitrary attributes of senders, messages, and receivers rather than traditional unicast or broadcast. We will be using SciClone to run an emulation of of a very large application network to measure the feasibility of our approach and the performance of our implementation.

[200412-01] Using Remote Memory with MMlib (C. Yue, 12/04)
This project is based on the existing Memory-Malleability Library (MMlib) framework that was developed by Richard T. Mills, Andreas Stathopoulos, and Dimitrios S. Nikolopoulos. Our objective is to add some mechanism in the current MMlib framework to provide application data from remote memory (via the network) instead of from local disk (via OS virtual memory) in a multiprogrammed environment. This approach is based on the observations that (1) network bandwidth is often higher than disk bandwidth, and (2) idle cycles and free memory resources are available in many computational environments.

[200105-01] Parallel Experimental Systems Software (N. Chrisochoides, K. Barker, 5/01)
Application driven research in mesh generation, parallel runtime systems, and problem solving environments for scientific computing.

[200207-01] Decoupling Method for Parallel Mesh Generation (L. Linardakis, D. Nave, 7/02)
The decoupling method is a domain decomposition based technique that uses an interface preprocessing approach to create large meshes in parallel. In the initial phase of this project we aim to create 2D meshes of order 109 elements.

[200202-01] Scalability Study of a Dynamic Load Balancing System for Adaptive Applications (A. Fedorov, 2/02)
In this project we will evaluate the scalability of a library designed to provide dynamic load balancing for asynchronous adaptive parallel applications. The load balancing library is designed to ease the costs of developing complex, dynamic load balancing applications as well as facilitate experimentation with various load balancing schemes. We will use mesh generation and refinement applications and compare the performance of the library with intrusive and explicit hard-wired load balancing methods.

Stochastic Load Balancing Methods Using the PREMA System (M. Balasubramaniam, N. Chrisochoides, 7/03)
Loops with variable iterate execution times need to be properly distributed among processors in order to obtain high performance on parallel systems. Over the years, a number of dynamic loop scheduling techniques have been proposed and used for different scientific applications. The intent of this work is to test the effectiveness of stochastic load balancing methods using the PREMA system (Parallel Runtime Enviromment for Mobile Applications). N-body simulations will be used as a case study.

[200011-01] Latency Tolerant Runtime Systems for Multi-Layer Parallel Architectures (K. Barker, A. Kot, 11/00)
Feasibility studies for the next generation of high-end parallel architectures indicate that these machines will consist of many layers of similar components for memory, processors, and interconnects with different levels of performance. This project is developing a runtime system to perform behavioral and structural analysis of algorithms and applications on parallel multi-layer architectures.

Runtime Support for Adaptive Out-of-Core Computations in MPI (D. Nikolopoulos, E. McCreedy, 4/03)
This project will implement a runtime library that will enable out-of-core (OOC) programs written with MPI and MPI I/O to adapt to CPU/memory load surges. The project will investigate methods based on control of buffer sizes and MPI's intercommunicators to enable adaptation in OOC programs.

Development and Testing of Advanced MPI Libraries (J. Squyres, B. Barrett, V. Sahay, 10/01)
Optimized Message Passing Interface (MPI) implementations are required for a broad range of scientific applications. This work focuses on extending the open source LAM implementation of MPI both to exploit the native message passing protocols of high speed switching fabrics such as Myrinet, and to explore scalability issues and requirements for large parallel jobs in commodity clusters. This is joint work with the Open Systems Lab at Indiana University.

Analysis, Simulation, and Monitoring Tools for Networks and Internets

[200307-04] Topology-Based Steering of Network Measurements (D. Bryan, 7/03)
This project is based on the observation that understanding network topology is key to understanding network performance. In particular, simple assumptions about network performance, such as WAN links being low-bandwidth, are useless with heterogeneous, irregular networks. Conversely, complete measurement of the performance between all pairs of hosts is unscalable. But a carefully managed, topology-based measurement system can successfully obtain the end-to-end network performance information needed by applications in a scalable, non-invasive manner.

[200310-01] Combining Passive Application Monitoring with Active Probes in a Network Monitoring Tool (M. Zangrilli, 10/03)
The goal of this project is to combine active and passive monitoring techniques to reduce the need for intrusive measurements without sacrificing accuracy. We are developing a bandwidth monitoring tool as part of the Wren network measurement system that will reduce the burden on the network by passively obtaining measurements from existing application traffic whenever possible, instead of actively probing the network.

Network Monitoring for the Remos System (B. Lowekamp, K. Blakely, S. Schmeelk, S. Small, L. Stratman, M. Zangrilli, 8/01)
This project uses SciClone to aid in the development of Remos, a system for monitoring and predicting network utilization. The cluster provides both a production environment to study the performance of Remos as well as the opportunity to generate synthetic workloads for controlled experiments.

Data Prefetching Between Web Servers and Proxy Caches (X. Chen, 11/00)
With accurate predictions of hyperlink paths in the future, web access latencies can be reduced through prefetching. Trace-driven simulations are being used to design and evaluate several prefetching techniques. SciClone is used to analyze surfing patterns in large web traces and to run trace-driven simulations of server-proxy prefetching models.

Graphics and Visualization

Parallel Methods for Graphics and Visualization (T. Crockett, 10/00)
This activity focuses on the development of parallel algorithms, methodologies, and software for computer graphics and scientific visualization. The emphasis is on techniques which can be employed at runtime to generate live visual output from parallel application programs. Principle areas of investigation include parallel surface and volume rendering algorithms, parallel libraries for graphics and visualization, performance modeling and analysis, networked delivery of image streams, remote interaction, and frameworks for developing application-specific user interfaces.

Data Management

Comparative Performance of SQL Implementations (J. Hines, R. Kaji, 3/03)
The purpose of this project is to compare, through timed testing, how efficiently pure SQL, PL/SQL, and SQL embedded in Java can process the same queries.

See also: