Parallel Computer Platforms Research Articles

Embedded Parallel Computing Platform for Real-Time Recognition of Power Quality Disturbance Based on Deep Network

Embedded Parallel Computing Platform for Real-Time Recognition of Power Quality Disturbance Based on Deep Network

The GlycoPaSER Prototype as a Real-Time N-Glycopeptide Identification Tool Based on the PaSER Parallel Computing Platform.

Real-time database searching allows for simpler and automated proteomics workflows as it eliminates technical bottlenecks in high-throughput experiments. Most importantly, it enables results-dependent acquisition (RDA), where search results can be used to guide data acquisition during acquisition. This is especially beneficial for glycoproteomics since the wide range of physicochemical properties of glycopeptides lead to a wide range of optimal acquisition parameters. We established here the GlycoPaSER prototype by extending the Parallel Search Engine in Real-time (PaSER) functionality for real-time glycopeptide identification from fragmentation spectra. Glycopeptide fragmentation spectra were decomposed into peptide and glycan moiety spectra using common N-glycan fragments. Each moiety was subsequently identified by a specialized algorithm running in real-time. GlycoPaSER can keep up with the rate of data acquisition for real-time analysis with similar performance to other glycoproteomics software and produces results that are in line with the literature reference data. The GlycoPaSER prototype presented here provides the first proof-of-concept for real-time glycopeptide identification that unlocks the future development of RDA technology to transcend data acquisition.

CWL-PLAS: Task Workflows Assisted by Data Science Cloud Platforms

The Common Workflow Language (CWL) is a platform-independent description language for the representation of data science workflows consisting of a set of tasks that interact with each other to perform scientific analysis. The tasks can be packaged as Linux containers. On the one hand, using containers ensures the reproducibility and portability of workflows. Still, on the other hand, it limits each task to exploiting, at most, the resources of the host where its container runs. In this paper, we propose CWL-PLAS, an extension of CWL that allows a task to instantiate and temporarily use a supporting cloud platform for parallel computing, which is specialized for the task’s activity. In this way, tasks can leverage the resources of multiple hosts in parallel, reducing the duration of the workflow. We implemented an open-source workflow manager that supports CWL-PLAS workflows and exploits a Kubernetes back-end. We used this workflow manager to evaluate the performance of CWL-PLAS in a couple of machine learning workflows.

A High Performance Global Routing Algorithm on Julia Parallel Computing Platform

o keep pace with the design requirements of Integrated Circuits (ICs), parallel processing is adopted. The path to be routed between two nodes may or may not be dependent on the previously routed paths. The solution requires careful attention in distributing the nets to be routed to different processors. Previous work on allocating the tasks to processors has been quite successful, reporting upto 3x improvement on 4 cores and 5x improvement on 8 core machine. The advantage of increasing the number of cores diminishes with each added processor and the challenge lies in being able to maintain the improvement per added core. The existing techniques of distributing the nets cannot provide additional advantage of using more than 8 cores. This paper improves the work on parallelizing global routing using a technique of balancing the load on the processors for better utilization of the resources. A relatively new budding platform Julia has been used which provides the ease of programming while maintaining the performance of the C language. Technique used in this paper has enabled to use 16 cores with routing solutions obtained in 0.8 minutes achieving 12.5 times speedup compared to sequential processing on a single core

Optimizing 2-opt-based heuristics on GPU for solving the single-row facility layout problem

The optimization of most combinatorial problems is NP-hard, which can be solved by heuristic algorithms to obtain approximately optimal solutions, especially for large-scale problems. Many heuristic algorithms can apply the 2-opt local search to find better solutions. In complete 2-opt local search, every valid neighboring solutions of swapping mechanism will be compared, which is very time consuming especially when a sequential algorithm is adopted. Nowadays, graphic processing units (GPUs) have evolved into powerful and flexible parallel computation platforms that have been widely used to accelerate the solving of NP-hard problems. Therefore, in this study, we focus on how to use GPUs to solve the single-row facility layout problem (SRFLP) with the 2-opt-based simulated annealing (SA) heuristic algorithm. As far as we know, this study is the first to solve SRFLP by using a GPU. After analyzing the fitness function and the move gains calculation between parent and child solutions, we propose a prefix-sum formula table to eliminate a large amount of replicated computation. To calculate move gains for the 2-opt local search during each iteration, many GPU threads are created and they construct and lookup the table in parallel. According to experimental results, if the prefix-sum formula table is not adopted, the GPU version outperforms the sequential CPU counterpart with the best speedup of 123. However, if the table is used in the GPU version, the best speedup can reach up to 3208. Because the proposed parallelization approach with the prefix-sum formula table is based on the features of the 2-opt local search and the SRFLP fitness function, it can be applied to any 2-opt-based heuristic algorithm for solving SRFLP even though the SA algorithm is used in this study.

Parallel rotor/stator interaction methods and steady/unsteady flow simulations of multi-row axial compressors

Rotor/stator interaction plays a crucial role in predicting the aerodynamic performance of high-speed multi-row axial compressor. Although mixing plane method has been widely used in steady simulations of multi-row single passage flows, the exchange of spanwise non-matched profiles of flow variable between adjacent blade rows is still unknown. Besides, quite little work was devoted to sliding mesh method for unsteady simulations of multi-row full-annulus compressor flows. In this work, mixing plane method and sliding mesh method were developed and presented, respectively, in the environment of parallel computer platforms. Particular attentions were paid to the accurate exchange of spanwise non-matched profiles of fluxes in mixing plane method, and the fast search algorithms of spanwise and circumferential non-matched meshes in sliding mesh method. The developed parallel rotor/stator interaction methods were validated with two multi-row high-speed axial compressors, i.e. NASA Stage 35 and first three stages of NASA 74A, both of which were calculated with steady flow simulation of one blade passage per row and unsteady flow simulation of full-annulus passages per row. Results highlighted that both the developed mixing plane method and sliding mesh method were able to predict the aerodynamic performance curve accurately and satisfy the interface flux conservation tightly. The flow wake and shock wave were observed to transfer continuously across rotor/stator interface with sliding mesh method in the unsteady full-annulus simulation. This work is valuable for the further development of advanced aerodynamic design and accurate performance prediction of multi-row axial compressors.

Asynchronous substructuring method with alternating local and global iterations

Until now, almost all investigations of asynchronous iterations within domain decomposition frameworks targeted methods of the parallel Schwarz type. A first, and sole, attempt to deal with a primal substructuring framework resulted in an asynchronous substructuring method where relaxation occurs simultaneously on the subdomains and on the interface between them, which therefore corresponds to a substructured relaxation scheme defined on the whole global domain. In this paper, we propose a Gauss–Seidel kind of improvement consisting of alternating between relaxation on the interface and relaxation on the subdomains, hence, always using the latest solutions in the subdomains when updating the solution on the interface, which is feasible at no additional cost. It turns out that one particular case of our general alternating relaxation scheme corresponds to an asynchronous substructuring method with iterations fully defined on the subdomains’ interface, and where only local Schur complements are involved. Practical performance evaluation on both standard Poisson’s and linear elasticity problems has been conducted using a multi-node parallel computational platform with up to 720 CPU cores.

IDCOS: optimization strategy for parallel complex expression computation on big data

Complex expressions are the basis of data analytics. To process complex expressions on big data efficiently, we developed a novel optimization strategy for parallel computation platforms such as Hadoop and Spark. We attempted to minimize the rounds of data repartition to achieve high performance. Aiming at this goal, we modeled the expression as a graph and developed a simplification algorithm for this graph. Based on the graph, we converted the round minimization problem into a graph decomposition problem and developed a linear algorithm for it. We also designed appropriated implementation for the optimization strategy. Extensive experimental results demonstrate that the proposed approach could optimize the computation of complex expressions effectively with small cost.

High Performance Computing in Parallel Electromagnetics Simulation Code suite ACE3P

A comprehensive set of parallel finite-element codes suite ACE3P (Advanced Computational Electromagnetics 3D Parallel) is developed by SLAC for multi-physics modeling of particle accelerators running on massively parallel computer platforms for high fidelity and high accuracy simulation. ACE3P enables rapid virtual prototyping of accelerator and RF component design, optimization and analysis. Advanced modeling capabilities have been facilitated by implementations of novel algorithms for numerical solvers. Code performance on state-of-the-art high performance computing (HPC) platforms for large-scale RF modeling in accelerator applications will be presented in this paper. All the simulations have been performed on the supercomputers at National Energy Research Computer Center (NERSC).

Deep Learning based Optimal Multimodal Fusion Framework for Intrusion Detection Systems for Healthcare Data

Data fusion is a multidisciplinary research area that involves different domains. It is used to attain minimum detection error probability and maximum reliability with the help of data retrieved from multiple healthcare sources. The generation of huge quantity of data from medical devices resulted in the formation of big data during which data fusion techniques become essential. Securing medical data is a crucial issue of exponentially-pacing computing world and can be achieved by Intrusion Detection Systems (IDS). In this regard, since singular-modality is not adequate to attain high detection rate, there is a need exists to merge diverse techniques using decision-based multimodal fusion process. In this view, this research article presents a new multimodal fusion-based IDS to secure the healthcare data using Spark. The proposed model involves decision-based fusion model which has different processes such as initialization, pre-processing, Feature Selection (FS) and multimodal classification for effective detection of intrusions. In FS process, a chaotic Butterfly Optimization (BO) algorithm called CBOA is introduced. Though the classic BO algorithm offers effective exploration, it fails in achieving faster convergence. In order to overcome this, i.e., to improve the convergence rate, this research work modifies the required parameters of BO algorithm using chaos theory. Finally, to detect intrusions, multimodal classifier is applied by incorporating three Deep Learning (DL)-based classification models. Besides, the concepts like Hadoop MapReduce and Spark were also utilized in this study to achieve faster computation of big data in parallel computation platform. To validate the outcome of the presented model, a series of experimentations was performed using the benchmark NSLKDDCup99 Dataset repository. The proposed model demonstrated its effective results on the applied dataset by offering the maximum accuracy of 99.21%, precision of 98.93% and detection rate of 99.59%. The results assured the betterment of the proposed model.

Many-Body Quantum Chemistry on Massively Parallel Computers.

The deployment of many-body quantum chemistry methods onto massively parallel high-performance computing (HPC) platforms is reviewed. The particular focus is on highly accurate methods that have become popular in predictive description of chemical phenomena, such as the coupled-cluster method. The account of relevant literature is preceded by a discussion of the modern and near-future HPC landscape and the relevant computational traits of the many-body methods, in their canonical and reduced-scaling formulations, that underlie the challenges in their HPC realization.

Inner product free iterative solution and elimination methods for linear systems of a three-by-three block matrix form

Large scale systems of algebraic equations are frequently solved by iterative solution methods, such as the conjugate gradient method for symmetric or a generalized conjugate gradient or generalized minimum residual method for nonsymmetric linear systems. In practice, to get an acceptable elapsed computing time when solving large scale problems, one shall use parallel computer platforms. However, such methods involve orthogonalization of search vectors which requires computation of many inner products and, hence, needs global communication of data, which will be costly in computer times. In this paper, we propose various inner product free methods, such as the Chebyshev acceleration method. We study the solution of linear systems arising from optimal control problems for PDEs, such as the edge element discretization of the time-periodic eddy current optimal control problem. Following a discretize-then-optimize scheme, the resulting linear system is of a three-by-three block matrix form. Various solution methods based on an approximate Schur complement and inner product free iterative solution methods for this linear system are analyzed and compared with an earlier used method for two-by-two block matrices with square blocks. The convergence properties and implementation details of the proposed methods are analyzed to show their effectiveness and practicality. Both serial and parallel numerical experiments are presented to further investigate the performance of the proposed methods compared with some other existing methods.

Realization of Parallelism in a Sequential Legacy ‘C’ Program

In the present era of high speed computation with the multicore and other parallel processors in the computational field, there are still some organizations which rely on their old software systems developed years ago, which over the time have been subjected to continous development by different developers. Even though these softwares persist with the old and little in use technology, they still work to satisfy the operational demands of the organizations and have kept them going in the competetive industry. These systems which have with time grown into legacy, embed the major business functionalities of the organization, which is but effort of years. Hence a methodology is required to rebuild the legacy system to make them suitable for execution on to the present computation systems. The paper discusses a research work, wherein work is done to realize points of latent parallelism in a sequentially executing legacy ‘C’ program which is initially restructured and the design information abstracted. A technique using finite state machine is proposed to identify tasks, events, processes and jobs in the program, which helps to locate functionally independent computational units in the program. Furthur using the slicing technique, slicing is performed to extract out the appropriate lines of codes defined by the slicing criteria, which assembled together form a functionality that can be executed in parallel with other extracted functional modules or computational units on any parallel computational platform.

Parallel Computing and Multicore Platform to Assess Electric Vehicle Hosting Capacity

The impact generated by the massive connection of electric vehicles (EVs) in the distribution networks needs to be evaluated through methodologies that consider the stochastic behavior of the EV connections. These methodologies must allow obtaining representative results of annual simulations, considering different charging scenarios, with high resolution of time steps. This kind of analysis implies a combinatorial explosion of case studies with a high computational cost. This article presents a methodology aimed at reducing the simulation time required to determine the EV hosting capacity (HC), understood as the maximum EV penetration level that a distribution network could host before affecting its operating limits, through the co-simulation of hardware and software. Monte Carlo methods and quasi-static time series simulation concepts are used to model the connection features of EVs. The simulation and implementation are developed using parallel computing, and MATLAB and OpenDSS software. The methodology is tested on the IEEE 123 nodes system which shows the dependence of the HC indicator with the location and type EV chargers. Finally, the benefits of parallel computing in the proposed methodology are exposed to get a significant reduction in the execution time to evaluate the HC in the IEEE 123 nodes system.

A New Concept of GPU-Accelerated On-Board Visual System for Aerial Vehicles

As technology advances, the level of intellectual ability and autonomy of the electronic-mechanical control system of modern aircrafts and spacecrafts is constantly growing, which helps to reduce the crew load and accident rate. But these helpful controlling systems are not perfect, under some unfavorable circumstances they get stuck or start to function unpredictably when faced with a much more complicated real situation than the developers expected, sometimes even lead to crashes. The concept of visual multichannel processing support of aircraft/spacecraft launch and landing as an additional element of automatic control loop for flight safety and reliability improvement is proposed and its advantages, feasibility and expediency are discussed and evaluated. The visual analyzers are very typical for the overwhelming majority of highly organized organisms (humans, animals, insects) as the most informative source of control of movement parameters, so they potentially can effectively improve the reliability of the entire embedded vehicle controlling system and, at the same time, their principal structure, implementation and further functioning are very similar and universal for real flight operation of different vehicles, which opens up great prospects for their application in engineering based on modern revolutionary achievements in the field of methodology and computational technologies for pattern recognition. In particular, it was shown that the required for real vehicles accuracy and productivity can be reached in case of developing the visual multichannel system, as an additional source of flight state information, on the base of NVIDIA Jetson embedded portable low power consuming Graphics processing unit (GPU)-accelerated massively parallel computational platform, providing CUDA and Artificial Intelligence data processing in real-time mode.

Parallel simulation of variably saturated soil water flows by fully implicit domain decomposition methods

The growing complexity of geometric models for the simulation of subsurface flows leads to the necessity of using the fully implicit method, due to its unconditionally stability with the relaxation of the time step size. In the paper, we introduce and study a parallel and scalable fully implicit solver for the simulation of variably saturated soil water flows. In the proposed approach, the flow problem is discretized in space and time based on a fully implicit, cell-centered finite volume scheme, where several different limiters are applied to guarantee the accuracy of the spatial discretization and an adaptive time-stepping technique for the temporal integration is designed to accelerate the simulation progress. And then the resultant nonlinear algebraic system, arising from the discretization of the variably saturated soil water flows, is solved by a family of Newton–Krylov methods under the domain decomposition framework. Numerical experiments on several standard benchmark tests are used to validate the stability of the fully implicit scheme with large time steps, and to examine the performance of the proposed domain decomposition solver on a parallel computer platform.

Molecular Dynamics Simulation of Gas Pipeline Steel Based on LAMMPS Parallel Computing Platform

Along with the advent of the fourth industrial revolution, the processing capacity of big is highly demanded in industry community. Compared with the traditional computing and programming model based on serial program, parallel computing is based on heterogeneous architecture and runs in a completely different way from CPU, which has greatly enhanced the processing speed of large-scale data sets. Therefore, parallel computing has been developing dramatically in recent years, and has been favored by all walks of life. In the future, large-scale data processing must belong to parallel computing. Set on LAMMPS (large-scale atomic/molecular massively parallel simulator) as an example, this paper described the technique of establishing parallel computing platform, and applied it on simulated test on fracture process of lamellar structure from gas-pipeline steel.

Molecular Dynamics Simulation of Gas Pipeline Steel Based on LAMMPS Parallel Computing Platform

Molecular Dynamics Simulation of Gas Pipeline Steel Based on LAMMPS Parallel Computing Platform

A Clustering Approach for Intrusion Detection with Big Data Processing on Parallel Computing Platform

Abstract— In recent years there is a growing number of attacks in the computer networks. Therefore, the use of a prevention mechanism is an inevitable need for security admins. Although firewalls are preferred as the first layer of protection, it is not sufficient for preventing lots of the attacks, especially from the insider attacks. Intrusion Detection Systems (IDSs) have emerged as an effective solution to these types of attacks. For increasing the efficiency of the IDS system, a dynamic solution, which can adapt itself and can detect new types of intrusions with a dynamic structure by the use of learning algorithms is mostly preferred. In previous years, some machine learning approaches are implemented in lots of IDSs. In the current position of artificial intelligence, most of the learning systems are transferred with the use of Deep Learning approaches due to its flexibility and the use of Big Data with high accuracy. In this paper, we propose a clustered approach to detect the intrusions in a network. Firstly, the system is trained with Deep Neural Network on a Big Data set by accelerating its performance with the use of CUDA architecture. Experimental results show that the proposed system has a very good accuracy rate and low runtime duration with the use of this parallel computation architecture. Additionally, the proposed system needs a relatively small duration for training the system

ADVANCED FEATURES OF NVIDIA KEPLER ARCHITECTURE AND PARALLEL COMPUTATION PLATFORM CUDA FOR DEVELOPING SCIENTIFIC COMPUTE-INTENSIVE APPLICATIONS

The paper describes additional features offered by new Kepler architecture of NVIDIA graphic processors, and their usage for creating high performance programs in a wide range of scientific compute-intensive applications. Recommendations are given for their use at realization of sci-tech computation algorithms by means of graphic processors. New capabilities of the parallel computation platform CUDA are also described, in particular, regarding a set of program development tool extensions for the Fortran, C and C++ languages. The extended capabilities make it possible to minimize the time of application development and to increase the programming productivity.