The field of Cyberinfrastructure (CI) has experienced rapid and exciting progress in enabling tec... more The field of Cyberinfrastructure (CI) has experienced rapid and exciting progress in enabling technologies, but much slower improvements in application codes, largely because of the ongoing disconnect between the CI and application research communities. At the University of Oklahoma, the OU Supercomputing Center for Education & Research (OSCER) is addressing this issue using a multipronged pedagogical strategy that combines workshops,
A computational method was developed to determine the correlation between permeability and the Fo... more A computational method was developed to determine the correlation between permeability and the Forchheimer (non-Darcy) coefficient and to investigate numerically the factors that affect this correlation. The method is based on the construction of a 2-D random pore network. The porous medium is represented as a network of cylindrical pipes with randomly generated size, orientation, and connectivity. Probability density functions
This paper describes the basic optimization methods applied to a Numerical Weather Prediction mod... more This paper describes the basic optimization methods applied to a Numerical Weather Prediction model, including improved cache memory usage and message hiding. The goal is to reduce the wall clock time for time critical weather forecasting and compute intensive research. The problem is put into perspective via a brief history of the type of computing hardware used for NWP applications
Although the field of High Performance Computing (HPC) has been evolv-ing rapidly, the developmen... more Although the field of High Performance Computing (HPC) has been evolv-ing rapidly, the development of standardized software systems – MPI, OpenMP, LAPACK, PETSc and others – has in principle made HPC accessible to a wider community. However, relatively few scientists and engineers take advantage of computational science and engineering (CSE) in their research, in part because of the considerable degree of sophistication about computing that HPC appears to re-quire. Yet the fundamental concepts of HPC are fairly straightforward and have remained relatively stable over time. These facts raise an important question: can scientists and engineers with relatively modest computing experience learn HPC concepts well enough to take advantage of them in their research? To answer this question, HPC educators must address several issues. First, what are the fundamental issues and concepts in CSE? Second, what are the fun-damental issues and concepts in HPC? Third, how can these ideas be express...
Proceedings of the Conference on Extreme Science and Engineering Discovery Environment Gateway to Discovery - XSEDE '13, 2013
ABSTRACT University-located computational and data storage resources are increasingly being aggre... more ABSTRACT University-located computational and data storage resources are increasingly being aggregated into shared university-level systems that are dubbed "condominium clusters." This session will describe the "Condo of Condos," a "consortium of the willing" that is working together to extend this condominium model both locally and across the consortium. By aggregating resources and expertise at the institutional level it will be possible to create distributed technical and support operations that provide faculty and students with methods to leverage a much larger and robust set of resources. Partnering with Internet2, the consortium will use the Innovation Platform to connect campuses and allow high performance interconnections through Science DMZs that can be built and broken down in ad-hoc ways using SDN. The ability to develop these network connections, coupled with the project's community building characteristics, will be a model for "team science" that can be extended and replicated nationally. By leveraging existing local and national computing resources (XSEDE and Open Science Grid), we will enable science and education deployments that are not possible today, making complex collaborations with computational needs routine. This project will be transformative for campus IT and will create a community of practitioners that shares resources, experience, and expertise to facilitate new knowledge and discovery.
INTRODUCTION Computing technologies are rapidly evolving, not only in speed and size, but also in... more INTRODUCTION Computing technologies are rapidly evolving, not only in speed and size, but also in structure, particu-larly with the rise of multicore and the anticipation of many-core. Concomitantly, the teaching of High Performance Computing (HPC) has become ...
... Teaching Cyberinfrastructure to Computing Novices ... not only in 17 academic departments at ... more ... Teaching Cyberinfrastructure to Computing Novices ... not only in 17 academic departments at OU, but also at 28 other institutions in 22 US states and territories as well as Mexico; recordings of these workshops are now available for free on the web as streaming video [4]. The ...
Commercial Applications for High-Performance Computing, 2001
ABSTRACT Fluid flow through porous materials is critical for understanding and predicting the beh... more ABSTRACT Fluid flow through porous materials is critical for understanding and predicting the behavior of systems as diverse in function and scale as hydrocarbon reservoirs, aquifers, filters, membrane separators and even catalytic converters. Recently, there have been calls to incorporate more physics in oil reservoir simulations, as well as to enhance computational capability through the use of High Performance Computing (HPC), in order to improve reservoir management. Accurate prediction of reservoir behavior depends on the physical properties of not only the fluid but also the underlying rock formation. Contemporary approaches to solving these flows involve simulation of only a single physical scale. We are currently developing HiMuST (Hierarchical Multiscale Simulator Technology), an integrated multiscale simulation system for flow through heterogeneous porous materials. HiMuST uses a hierarchy of simulation codes to address the issue of rock property characterization at the pore scale and can self-adjust according to available input data. At the microscopic scale, HiMuST employs the Lattice Boltzmann Method, based on magnetic resonance digitizations of actual rock samples. At the mesoscopic scale, a stochastic model represents a pore network as a randomly generated skeleton of cylindrical pipes, based on physical characteristics determined by the microscopic simulation. We present computational and computer science issues involved in the HPC implementation of the codes and in integrating them into a seamless simulation system. Issues such as portability, scalability, efficiency and extensibility of the final product are also discussed, as well as the numerical methods implemented at each scale. Example simulation results are presented.
2013 IEEE International Conference on Cluster Computing (CLUSTER), 2013
ABSTRACT In the era of Big Data, research productivity can be highly sensitive to the availabilit... more ABSTRACT In the era of Big Data, research productivity can be highly sensitive to the availability of large scale, long term storage. Unfortunately, most mass storage systems are prohibitively expensive at scales appropriate for individual institutions rather than national centers. Furthermore, many researchers won't adopt any centralized technology that they perceive as more expensive than what their research teams could do on their own. This poster examines a combination of business model and technology that addresses these concerns in a comprehensive way, distributing the costs among a funding agency, the institution and the research teams, thereby reducing the challenges faced by each.
Proceedings of the 2010 TeraGrid Conference on - TG '10, 2010
... Skylar Thompson Genome Sciences Department University of Washington Seattle, WA 98195skylar@c... more ... Skylar Thompson Genome Sciences Department University of Washington Seattle, WA 98195skylar@cs.earlham.edu ... 2.1 Workshop Content The general format of the workshops is to start on Sun-day evening, with introductory presentations and a tour of the local HPC facility if ...
ABSTRACT The goal of this study was to determine laminar stagnation pressure loss coefficients fo... more ABSTRACT The goal of this study was to determine laminar stagnation pressure loss coefficients for circular ducts in which flow encounters a planar bifurcation. Flow conditions and pressure losses in these laminar bifurcations are of interest in microfluidic devices, in porous media, and in other networks of small ducts or pores. Until recently, bifurcation geometries had been studied almost exclusively for turbulent flow, which is often found in fluid supply and drain systems. Recently, pressure loss coefficients from simulations of a few arbitrary bifurcation geometries in two-dimensions have been published — the present study describes the extension of these two-dimensional simulations to three-dimensional circular ducts. The pressure loss coefficients determined in this study are intended to allow realistic simulation of existing laminar flow networks or the design of these networks. This study focused on a single inlet duct with two outlet ducts, which were allowed to vary in diameter, flow fraction, and angle — all relative to the inlet duct. All ducts considered in this study were circular with their axes in a common plane. Laminar stagnation pressure loss coefficients were determined by simulating incompressible flow through 475 different geometries and flow condition combinations. In all cases, the flow was laminar in the inlet and outlet ducts with a Reynolds number of 15 in the inlet duct. Simulations of the dividing flow geometries were done using FLUENT and a custom written computer code, which automated the process of creating the three-dimensional flow geometries. The outputs, pressure and velocity distributions at the inlet and outlets, were averaged over the circular ducts and then used to calculate pressure loss coefficients for each of the geometries and flow fraction scenarios simulated. The results for loss coefficient for the geometries considered ranged from 2.0 to 70. The loss coefficient for any geometry increased significantly as the outlet flow fraction increased. A consistent increase in loss coefficient was also observed as a function of decreasing outlet duct diameter. Less significant variation of the loss coefficient was observed as a function of the angles of the outlet ducts.
Parallel computing is increasingly a requirement of Computational Science and Engineering (CSE) s... more Parallel computing is increasingly a requirement of Computational Science and Engineering (CSE) software because problems of interest are very large and hardware systems are becoming parallel through multicore technologies. Inexperienced programmers-non-computer scientists with one semester to a few years of programming experience-are crucial to CSE software development because academic research teams rely on them as application developers. For this group,
International Journal of Computer Applications in Technology, 2005
ABSTRACT In the solution of streamflow problems, fully distributed hydrologic models (DHMs), whic... more ABSTRACT In the solution of streamflow problems, fully distributed hydrologic models (DHMs), which are based directly on governing equations, offer distinct advantages over conceptual rainfall-runoff models, which are derived from empirical observations. However, the primary problem associated with DHMs is that they consume more computational resources than other models, and as a result, they have not been as popular as their capabilities would imply. A parallel DHM software system for solving streamflow prediction problems has been implemented and analysed, and an investigation has been conducted of: the efficiency and scalability of the algorithm; load balancing among processors; interprocessor communication; disk performance. The load balancing algorithms show great promise for the kind of problem addressed. The software exhibits substantial parallel speedup, but the degree of speedup is limited by I/O costs.
Computer Methods and Programs in Biomedicine, 1990
This paper describes software developed to analyze the temporal and spatial aspects of jaw moveme... more This paper describes software developed to analyze the temporal and spatial aspects of jaw movements, the speed, direction and magnitude of these movements in three dimensions, and the associated electromyograms (EMGs) of the masticatory muscles. Data from a subject chewing gum illustrates the software application. Manipulations of the stored digitized data are described. Temporal aspects of each chewing cycle are defined by specified delimiters for the beginning and ending of jaw opening and closing; spatial aspects are quantified by measurements derived from movement trajectories; and peak and mean velocities are derived from the velocity profiles. EMGs are digitized, rectified, averaged and referenced to jaw movement. Onset latencies, burst durations, rise times to peak activity, time of occurrence of peak activity, and amplitudes of peak and mean activities are measured from the EMGs. This detailed kinematic/EMG analysis provides previously inaccessible information about jaw muscles and the movements they control.
Assembly of viruses that have hundreds of subunits or folding of proteins that have hundreds of a... more Assembly of viruses that have hundreds of subunits or folding of proteins that have hundreds of amino acidscomplex biological reactions-are often spontaneous and rapid. Here, we examine the complete set of intermediates available for the assembly of a hypothetical viruslike particle and the connectivity between these intermediates in a graph-theory-inspired study. Using a build-up procedure, assuming ideal geometry, we enumerated the complete set of 2,423,313 species for formation of an icosahedron from 30 dimeric subunits. Stability of each n-subunit intermediate was defined by the number of contacts between subunits. The probability of forming an intermediate was based on the number of paths to it from its precedecessors. When defining population subsets predicted to have the greatest impact on assembly, both stability-and probability-based criteria select a small group of compact and degenerate species; ergo, only a few hundred intermediates make a measurable contribution to assembly. Though the number of possible intermediates grows combinatorially with the number of subunits in the capsid, the number of intermediates that make a significant contribution to the reaction grows by a much smaller function, a result that may contribute to our understanding of assembly and folding reactions.
The field of Cyberinfrastructure (CI) has experienced rapid and exciting progress in enabling tec... more The field of Cyberinfrastructure (CI) has experienced rapid and exciting progress in enabling technologies, but much slower improvements in application codes, largely because of the ongoing disconnect between the CI and application research communities. At the University of Oklahoma, the OU Supercomputing Center for Education & Research (OSCER) is addressing this issue using a multipronged pedagogical strategy that combines workshops,
A computational method was developed to determine the correlation between permeability and the Fo... more A computational method was developed to determine the correlation between permeability and the Forchheimer (non-Darcy) coefficient and to investigate numerically the factors that affect this correlation. The method is based on the construction of a 2-D random pore network. The porous medium is represented as a network of cylindrical pipes with randomly generated size, orientation, and connectivity. Probability density functions
This paper describes the basic optimization methods applied to a Numerical Weather Prediction mod... more This paper describes the basic optimization methods applied to a Numerical Weather Prediction model, including improved cache memory usage and message hiding. The goal is to reduce the wall clock time for time critical weather forecasting and compute intensive research. The problem is put into perspective via a brief history of the type of computing hardware used for NWP applications
Although the field of High Performance Computing (HPC) has been evolv-ing rapidly, the developmen... more Although the field of High Performance Computing (HPC) has been evolv-ing rapidly, the development of standardized software systems – MPI, OpenMP, LAPACK, PETSc and others – has in principle made HPC accessible to a wider community. However, relatively few scientists and engineers take advantage of computational science and engineering (CSE) in their research, in part because of the considerable degree of sophistication about computing that HPC appears to re-quire. Yet the fundamental concepts of HPC are fairly straightforward and have remained relatively stable over time. These facts raise an important question: can scientists and engineers with relatively modest computing experience learn HPC concepts well enough to take advantage of them in their research? To answer this question, HPC educators must address several issues. First, what are the fundamental issues and concepts in CSE? Second, what are the fun-damental issues and concepts in HPC? Third, how can these ideas be express...
Proceedings of the Conference on Extreme Science and Engineering Discovery Environment Gateway to Discovery - XSEDE '13, 2013
ABSTRACT University-located computational and data storage resources are increasingly being aggre... more ABSTRACT University-located computational and data storage resources are increasingly being aggregated into shared university-level systems that are dubbed "condominium clusters." This session will describe the "Condo of Condos," a "consortium of the willing" that is working together to extend this condominium model both locally and across the consortium. By aggregating resources and expertise at the institutional level it will be possible to create distributed technical and support operations that provide faculty and students with methods to leverage a much larger and robust set of resources. Partnering with Internet2, the consortium will use the Innovation Platform to connect campuses and allow high performance interconnections through Science DMZs that can be built and broken down in ad-hoc ways using SDN. The ability to develop these network connections, coupled with the project's community building characteristics, will be a model for "team science" that can be extended and replicated nationally. By leveraging existing local and national computing resources (XSEDE and Open Science Grid), we will enable science and education deployments that are not possible today, making complex collaborations with computational needs routine. This project will be transformative for campus IT and will create a community of practitioners that shares resources, experience, and expertise to facilitate new knowledge and discovery.
INTRODUCTION Computing technologies are rapidly evolving, not only in speed and size, but also in... more INTRODUCTION Computing technologies are rapidly evolving, not only in speed and size, but also in structure, particu-larly with the rise of multicore and the anticipation of many-core. Concomitantly, the teaching of High Performance Computing (HPC) has become ...
... Teaching Cyberinfrastructure to Computing Novices ... not only in 17 academic departments at ... more ... Teaching Cyberinfrastructure to Computing Novices ... not only in 17 academic departments at OU, but also at 28 other institutions in 22 US states and territories as well as Mexico; recordings of these workshops are now available for free on the web as streaming video [4]. The ...
Commercial Applications for High-Performance Computing, 2001
ABSTRACT Fluid flow through porous materials is critical for understanding and predicting the beh... more ABSTRACT Fluid flow through porous materials is critical for understanding and predicting the behavior of systems as diverse in function and scale as hydrocarbon reservoirs, aquifers, filters, membrane separators and even catalytic converters. Recently, there have been calls to incorporate more physics in oil reservoir simulations, as well as to enhance computational capability through the use of High Performance Computing (HPC), in order to improve reservoir management. Accurate prediction of reservoir behavior depends on the physical properties of not only the fluid but also the underlying rock formation. Contemporary approaches to solving these flows involve simulation of only a single physical scale. We are currently developing HiMuST (Hierarchical Multiscale Simulator Technology), an integrated multiscale simulation system for flow through heterogeneous porous materials. HiMuST uses a hierarchy of simulation codes to address the issue of rock property characterization at the pore scale and can self-adjust according to available input data. At the microscopic scale, HiMuST employs the Lattice Boltzmann Method, based on magnetic resonance digitizations of actual rock samples. At the mesoscopic scale, a stochastic model represents a pore network as a randomly generated skeleton of cylindrical pipes, based on physical characteristics determined by the microscopic simulation. We present computational and computer science issues involved in the HPC implementation of the codes and in integrating them into a seamless simulation system. Issues such as portability, scalability, efficiency and extensibility of the final product are also discussed, as well as the numerical methods implemented at each scale. Example simulation results are presented.
2013 IEEE International Conference on Cluster Computing (CLUSTER), 2013
ABSTRACT In the era of Big Data, research productivity can be highly sensitive to the availabilit... more ABSTRACT In the era of Big Data, research productivity can be highly sensitive to the availability of large scale, long term storage. Unfortunately, most mass storage systems are prohibitively expensive at scales appropriate for individual institutions rather than national centers. Furthermore, many researchers won't adopt any centralized technology that they perceive as more expensive than what their research teams could do on their own. This poster examines a combination of business model and technology that addresses these concerns in a comprehensive way, distributing the costs among a funding agency, the institution and the research teams, thereby reducing the challenges faced by each.
Proceedings of the 2010 TeraGrid Conference on - TG '10, 2010
... Skylar Thompson Genome Sciences Department University of Washington Seattle, WA 98195skylar@c... more ... Skylar Thompson Genome Sciences Department University of Washington Seattle, WA 98195skylar@cs.earlham.edu ... 2.1 Workshop Content The general format of the workshops is to start on Sun-day evening, with introductory presentations and a tour of the local HPC facility if ...
ABSTRACT The goal of this study was to determine laminar stagnation pressure loss coefficients fo... more ABSTRACT The goal of this study was to determine laminar stagnation pressure loss coefficients for circular ducts in which flow encounters a planar bifurcation. Flow conditions and pressure losses in these laminar bifurcations are of interest in microfluidic devices, in porous media, and in other networks of small ducts or pores. Until recently, bifurcation geometries had been studied almost exclusively for turbulent flow, which is often found in fluid supply and drain systems. Recently, pressure loss coefficients from simulations of a few arbitrary bifurcation geometries in two-dimensions have been published — the present study describes the extension of these two-dimensional simulations to three-dimensional circular ducts. The pressure loss coefficients determined in this study are intended to allow realistic simulation of existing laminar flow networks or the design of these networks. This study focused on a single inlet duct with two outlet ducts, which were allowed to vary in diameter, flow fraction, and angle — all relative to the inlet duct. All ducts considered in this study were circular with their axes in a common plane. Laminar stagnation pressure loss coefficients were determined by simulating incompressible flow through 475 different geometries and flow condition combinations. In all cases, the flow was laminar in the inlet and outlet ducts with a Reynolds number of 15 in the inlet duct. Simulations of the dividing flow geometries were done using FLUENT and a custom written computer code, which automated the process of creating the three-dimensional flow geometries. The outputs, pressure and velocity distributions at the inlet and outlets, were averaged over the circular ducts and then used to calculate pressure loss coefficients for each of the geometries and flow fraction scenarios simulated. The results for loss coefficient for the geometries considered ranged from 2.0 to 70. The loss coefficient for any geometry increased significantly as the outlet flow fraction increased. A consistent increase in loss coefficient was also observed as a function of decreasing outlet duct diameter. Less significant variation of the loss coefficient was observed as a function of the angles of the outlet ducts.
Parallel computing is increasingly a requirement of Computational Science and Engineering (CSE) s... more Parallel computing is increasingly a requirement of Computational Science and Engineering (CSE) software because problems of interest are very large and hardware systems are becoming parallel through multicore technologies. Inexperienced programmers-non-computer scientists with one semester to a few years of programming experience-are crucial to CSE software development because academic research teams rely on them as application developers. For this group,
International Journal of Computer Applications in Technology, 2005
ABSTRACT In the solution of streamflow problems, fully distributed hydrologic models (DHMs), whic... more ABSTRACT In the solution of streamflow problems, fully distributed hydrologic models (DHMs), which are based directly on governing equations, offer distinct advantages over conceptual rainfall-runoff models, which are derived from empirical observations. However, the primary problem associated with DHMs is that they consume more computational resources than other models, and as a result, they have not been as popular as their capabilities would imply. A parallel DHM software system for solving streamflow prediction problems has been implemented and analysed, and an investigation has been conducted of: the efficiency and scalability of the algorithm; load balancing among processors; interprocessor communication; disk performance. The load balancing algorithms show great promise for the kind of problem addressed. The software exhibits substantial parallel speedup, but the degree of speedup is limited by I/O costs.
Computer Methods and Programs in Biomedicine, 1990
This paper describes software developed to analyze the temporal and spatial aspects of jaw moveme... more This paper describes software developed to analyze the temporal and spatial aspects of jaw movements, the speed, direction and magnitude of these movements in three dimensions, and the associated electromyograms (EMGs) of the masticatory muscles. Data from a subject chewing gum illustrates the software application. Manipulations of the stored digitized data are described. Temporal aspects of each chewing cycle are defined by specified delimiters for the beginning and ending of jaw opening and closing; spatial aspects are quantified by measurements derived from movement trajectories; and peak and mean velocities are derived from the velocity profiles. EMGs are digitized, rectified, averaged and referenced to jaw movement. Onset latencies, burst durations, rise times to peak activity, time of occurrence of peak activity, and amplitudes of peak and mean activities are measured from the EMGs. This detailed kinematic/EMG analysis provides previously inaccessible information about jaw muscles and the movements they control.
Assembly of viruses that have hundreds of subunits or folding of proteins that have hundreds of a... more Assembly of viruses that have hundreds of subunits or folding of proteins that have hundreds of amino acidscomplex biological reactions-are often spontaneous and rapid. Here, we examine the complete set of intermediates available for the assembly of a hypothetical viruslike particle and the connectivity between these intermediates in a graph-theory-inspired study. Using a build-up procedure, assuming ideal geometry, we enumerated the complete set of 2,423,313 species for formation of an icosahedron from 30 dimeric subunits. Stability of each n-subunit intermediate was defined by the number of contacts between subunits. The probability of forming an intermediate was based on the number of paths to it from its precedecessors. When defining population subsets predicted to have the greatest impact on assembly, both stability-and probability-based criteria select a small group of compact and degenerate species; ergo, only a few hundred intermediates make a measurable contribution to assembly. Though the number of possible intermediates grows combinatorially with the number of subunits in the capsid, the number of intermediates that make a significant contribution to the reaction grows by a much smaller function, a result that may contribute to our understanding of assembly and folding reactions.
Uploads
Papers by Henry Neeman