Search
displaying 1000 best results Help
Search results for: parallel mpi i/o extension
-
A Solution to Image Processing with Parallel MPI I/O and Distributed NVRAM Cache
PublicationThe paper presents a new approach to parallel image processing using byte addressable, non-volatile memory (NVRAM). We show that our custom built MPI I/O implementation of selected functions that use a distributed cache that incorporates NVRAMs located in cluster nodes can be used for efficient processing of large images. We demonstrate performance benefits of such a solution compared to a traditional implementation without NVRAM...
-
A Parallel MPI I/O Solution Supported by Byte-addressable Non-volatile RAM Distributed Cache
PublicationWhile many scientific, large-scale applications are data-intensive, fast and efficient I/O operations have become of key importance for HPC environments. We propose an MPI I/O extension based on in-system distributed cache with data located in Non-volatile Random Access Memory (NVRAM) available in each cluster node. The presented architecture makes effective use of NVRAM properties such as persistence and byte-level access behind...
-
A Fail-Safe NVRAM Based Mechanism for Efficient Creation and Recovery of Data Copies in Parallel MPI Applications
PublicationThe paper presents a fail-safe NVRAM based mechanism for creation and recovery of data copies during parallel MPI application runtime. Specifically, we target a cluster environment in which each node has an NVRAM installed in it. Our previously developed extension to the MPI I/O API can take advantage of NVRAM regions in order to provide an NVRAM based cache like mechanism to significantly speed up I/O operations and allow to preload...
-
Checkpointing of Parallel MPI Applications using MPI One-sided API with Support for Byte-addressable Non-volatile RAM
PublicationThe increasing size of computational clusters results in an increasing probability of failures, which in turn requires application checkpointing in order to survive those failures. Traditional checkpointing requires data to be copied from application memory into persistent storage medium, which increases application execution time as it is usually done in a separate step. In this paper we propose to use emerging byte-addressable...
-
New user-guided and ckpt-based checkpointing libraries for parallel MPI applications
PublicationPraca prezentuje szczególy projektowe i implementacyjne jak również wyniki wydajnościowe dwóch nowych bibliotek checkpointingu opracowanych przez autorów dla równoległych aplikacji MPI. Pierwsz biblioteka, tzw. user-guided wymaga od programisty dostarczenia funkcji pakujących i rozpakowujących stan procesu, ale dostarcza łatwego w użyciu API z wykorzystaniem stałych MPI. Wykorzystuje funkcje I/O MPI-2 lub dedykowany proces master...
-
Performance Assessment of Using Docker for Selected MPI Applications in a Parallel Environment Based on Commodity Hardware
PublicationIn the paper, we perform detailed performance analysis of three parallel MPI applications run in a parallel environment based on commodity hardware, using Docker and bare-metal configurations. The testbed applications are representative of the most typical parallel processing paradigms: master–slave, geometric Single Program Multiple Data (SPMD) as well as divide-and-conquer and feature characteristic computational and communication...
-
Portable parallel simulator using MPI for 2D and 3D domains: design and performance testing
PublicationW artykule prezentujemy szczegóły projektowo-implementacyjne naszego modularnego kodu symulacyjnego z wykorzystaniem MPI, w tym nakładaniem obliczeń i komunikacji. Podkreślamy modularność naszej implementacji pozwalającą na łatwą adaptację kodu dla innych zasotosowań. Prezentujemy związek pomiędzy przyspieszeniem obliczeń, rozmiarem i kształtami trójwymiarowych domen z różnymi stosunkami liczby węzłów aktualizowanych przez procesor...
-
Paweł Czarnul dr hab. inż.
PeoplePaweł Czarnul obtained a D.Sc. degree in computer science in 2015, a Ph.D. in computer science granted by a council at the Faculty of Electronics, Telecommunications and Informatics, Gdansk University of Technology in 2003. His research interests include:parallel and distributed processing including clusters, accelerators, coprocessors; distributed information systems; architectures of distributed systems; programming mobile devices....
-
Multi-agent large-scale parallel crowd simulation
PublicationThis paper presents design, implementation and performance results of a new modular, parallel, agent-based and large scale crowd simulation environment. A parallel application, implemented with C and MPI, was implemented and run in this parallel environment for simulation and visualization of an evacuation scenario at Gdansk University of Technology, Poland and further in the area of districts of Gdansk. The application uses a...
-
Parallelization of Compute Intensive Applications into Workflows based on Services in BeesyCluster
PublicationThe paper presents an approach for modeling, optimization and execution of workflow applications based on services that incorporates both service selection and partitioning of input data for parallel processing by parallel workflow paths. A compute-intensive workflow application for parallel integration is presented. An impact of the input data partitioning on the scalability is presented. The paper shows a comparison of the theoretical...
-
Simulation of parallel similarity measure computations for large data sets
PublicationThe paper presents our approach to implementation of similarity measure for big data analysis in a parallel environment. We describe the algorithm for parallelisation of the computations. We provide results from a real MPI application for computations of similarity measures as well as results achieved with our simulation software. The simulation environment allows us to model parallel systems of various sizes with various components...
-
Performance and Power-Aware Modeling of MPI Applications for Cluster Computing
PublicationThe paper presents modeling of performance and power consumption when running parallel applications on modern cluster-based systems. The model includes basic so-called blocks representing either computations or communication. The latter includes both point-to-point and collective communication. Real measurements were performed using MPI applications and routines run on three different clusters with both Infiniband and Gigabit Ethernet...
-
NVRAM as Main Storage of Parallel File System
PublicationModern cluster environments' main trouble used to be lack of computational power provided by CPUs and GPUs, but recently they suffer more and more from insufficient performance of input and output operations. Apart from better network infrastructure and more sophisticated processing algorithms, a lot of solutions base on emerging memory technologies. This paper presents evaluation of using non-volatile random-access memory as a...
-
An innovative method of measuring the extension of the piston rod in hydraulic cylinders, especially large ones used in the shipbuilding and offshore industry
PublicationThe article presents the results of selected works related to the wider subject of research conducted at the Faculty of Mechanical Engineering and Shipbuilding at the Gdańsk University of Technology, regarding designing various on board devices with hydraulic drive for ships and other offshore facilities. One of the commonly used these mechanisms are hydraulic actuators with the measurement of the piston rod extension. The issue...
-
Modeling energy consumption of parallel applications
PublicationThe paper presents modeling and simulation of energy consumption of two types of parallel applications: geometric Single Program Multiple Data (SPMD) and divide-and-conquer (DAC). Simulation is performed in a new MERPSYS environment. Model of an application uses the Java language with extension representing message exchange between processes working in parallel. Simulation is performed by running threads representing distinct process...
-
Parallel Programming for Modern High Performance Computing Systems
PublicationIn view of the growing presence and popularity of multicore and manycore processors, accelerators, and coprocessors, as well as clusters using such computing devices, the development of efficient parallel applications has become a key challenge to be able to exploit the performance of such systems. This book covers the scope of parallel programming for modern high performance computing systems. It first discusses selected and...
-
Multi-GPU-powered UNRES package for physics-based coarse-grained simulations of structure, dynamics, and thermodynamics of protein systems at biological size- and timescales
PublicationCoarse-grained models are nowadays extensively used in biomolecular simulations owing to the tremendous extension of size- and time-scale of simulations. The physics-based UNRES (UNited RESidue) model of proteins developed in our laboratory has only two interaction sites per amino-acid residue (united peptide groups and united side chains) and implicit solvent. However, owing to rigorous physics-based derivation, which enabled...
-
BC-MPI: running an mpi application on multiple clusters with beesycluster connectivity
PublicationW artykule zaproponowano nowy pakiet BC-MPI, który umożliwiauruchomienie aplikacji MPI na wielu klastrach z różnymi implementacjami MPI. Wykorzystuje dedykowane implementacje MPIdo komunikacji wewnątrz klastrów oraz tryb MPI THREAD MULTIPLE dokomunikacji pomiędzy klastrami w dodatkowych wątkach aplikacji MPI. Ponadto, aplikacja BC-MPI może być automatycznie skompilowanai uruchomiona przez warstwę pośrednią BeesyCluster. BeesyClusterumożliwia...
-
MERPSYS: An environment for simulation of parallel application execution on large scale HPC systems
PublicationIn this paper we present a new environment called MERPSYS that allows simulation of parallel application execution time on cluster-based systems. The environment offers a modeling application using the Java language extended with methods representing message passing type communication routines. It also offers a graphical interface for building a system model that incorporates various hardware components such as CPUs, GPUs, interconnects...
-
Distributed NVRAM Cache – Optimization and Evaluation with Power of Adjacency Matrix
PublicationIn this paper we build on our previously proposed MPI I/O NVRAM distributed cache for high performance computing. In each cluster node it incorporates NVRAMs which are used as an intermediate cache layer between an application and a file for fast read/write operations supported through wrappers of MPI I/O functions. In this paper we propose optimizations of the solution including handling of write requests with a synchronous mode,...
-
Investigation into MPI All-Reduce Performance in a Distributed Cluster with Consideration of Imbalanced Process Arrival Patterns
PublicationThe paper presents an evaluation of all-reduce collective MPI algorithms for an environment based on a geographically-distributed compute cluster. The testbed was split into two sites: CI TASK in Gdansk University of Technology and ICM in University of Warsaw, located about 300 km from each other, both connected by a fast optical fiber Ethernet-based 100 Gbps network (900 km part of the PIONIER backbone). Each site hosted a set...
-
Parallel multithread computing for spectroscopic analysis in optical coherence tomography
PublicationSpectroscopic Optical Coherence Tomography (SOCT) is an extension of Optical Coherence Tomography (OCT). It allows gathering spectroscopic information from individual scattering points inside the sample. It is based on time-frequency analysis of interferometric signals. Such analysis requires calculating hundreds of Fourier transforms while performing a single A-scan. Additionally, further processing of acquired spectroscopic information...
-
Process arrival pattern aware algorithms for acceleration of scatter and gather operations
PublicationImbalanced process arrival patterns (PAPs) are ubiquitous in many parallel and distributed systems, especially in HPC ones. The collective operations, e.g. in MPI, are designed for equal process arrival times (PATs), and are not optimized for deviations in their appearance. We propose eight new PAP-aware algorithms for the scatter and gather operations. They are binomial or linear tree adaptations introducing additional process...
-
Parallelization of Selected Algorithms on Multi-core CPUs, a Cluster and in a Hybrid CPU+Xeon Phi Environment
PublicationIn the paper we present parallel implementations as well as execution times and speed-ups of three different algorithms run in various environments such as on a workstation with multi-core CPUs and a cluster. The parallel codes, implementing the master-slave model in C+MPI, differ in computation to communication ratios. The considered problems include: a genetic algorithm with various ratios of master processing time to communication...
-
Minimizing Distribution and Data Loading Overheads in Parallel Training of DNN Acoustic Models with Frequent Parameter Averaging
PublicationIn the paper we investigate the performance of parallel deep neural network training with parameter averaging for acoustic modeling in Kaldi, a popular automatic speech recognition toolkit. We describe experiments based on training a recurrent neural network with 4 layers of 800 LSTM hidden states on a 100-hour corpora of annotated Polish speech data. We propose a MPI-based modification of the training program which minimizes the...
-
KernelHive: a new workflow-based framework for multilevel high performance computing using clusters and workstations with CPUs and GPUs
PublicationThe paper presents a new open-source framework called KernelHive for multilevel parallelization of computations among various clusters, cluster nodes, and finally, among both CPUs and GPUs for a particular application. An application is modeled as an acyclic directed graph with a possibility to run nodes in parallel and automatic expansion of nodes (called node unrolling) depending on the number of computation units available....
-
Two Stage SVM and kNN Text Documents Classifier
PublicationThe paper presents an approach to the large scale text documents classification problem in parallel environments. A two stage classifier is proposed, based on a combination of k-nearest neighbors and support vector machines classification methods. The details of the classifier and the parallelisation of classification, learning and prediction phases are described. The classifier makes use of our method named one-vs-near. It is...
-
A multithreaded CUDA and OpenMP based power‐aware programming framework for multi‐node GPU systems
PublicationIn the paper, we have proposed a framework that allows programming a parallel application for a multi-node system, with one or more GPUs per node, using an OpenMP+extended CUDA API. OpenMP is used for launching threads responsible for management of particular GPUs and extended CUDA calls allow to manage CUDA objects, data and launch kernels. The framework hides inter-node MPI communication from the programmer who can benefit from...
-
Improving Clairvoyant: reduction algorithm resilient to imbalanced process arrival patterns
PublicationThe Clairvoyant algorithm proposed in “A novel MPI reduction algorithm resilient to imbalances in process arrival times” was analyzed, commented and improved. The comments concern handling certain edge cases in the original pseudocode and description, i.e., adding another state of a process, improved cache friendliness more precise complexity estimations and some other issues improving the robustness of the algorithm implementation....
-
Use of ICT infrastructure for teaching HPC
PublicationIn this paper we look at modern ICT infrastructure as well as curriculum used for conducting a contemporary course on high performance computing taught over several years at the Faculty of Electronics Telecommunications and Informatics, Gdansk University of Technology, Poland. We describe the infrastructure in the context of teaching parallel programming at the cluster level using MPI, node level using OpenMP and CUDA. We present...
-
Improving Effectiveness of SVM Classifier for Large Scale Data
PublicationThe paper presents our approach to SVM implementation in parallel environment. We describe how classification learning and prediction phases were pararellised. We also propose a method for limiting the number of necessary computations during classifier construction. Our method, named one-vs-near, is an extension of typical one-vs-all approach that is used for binary classifiers to work with multiclass problems. We perform experiments...
-
Variable-structure algorithm for identification of quasi-periodically varying systems
PublicationThe paper presents a variable-structure version of a generalized notchfiltering (GANF) algorithm. Generalized notch filters are used for identification of quasi-periodically varying dynamic systems and can be considered an extension, to the system case, of classical adaptive notch filters. The proposed algorithm is a cascade of two GANF filters: a multiple-frequency "precise" filter bank, used for precise system tracking, and a...
-
Locally Adaptive Cooperative Kalman Smoothing and Its Application to Identification of Nonstationary Stochastic Systems
PublicationOne of the central problems of the stochastic approximation theory is the proper adjustment of the smoothing algorithm to the unknown, and possibly time-varying, rate and mode of variation of the estimated signals/parameters. In this paper we propose a novel locally adaptive parallel estimation scheme which can be used to solve the problem of fixed-interval Kalman smoothing in the presence of model uncertainty. The proposed solution...
-
FPGA Acceleration of Matrix-Assembly Phase of RWG-Based MoM
PublicationIn this letter, the field-programmable-gate-array accelerated implementation of matrix-assembly phase of the method of moments (MoM) is presented. The solution is based on a discretization of the frequency-domain mixed potential integral equation using the Rao-Wilton-Glisson basis functions and their extension to wire-to-surface junctions. To take advantage of the given hardware resources (i.e., Xilinx Alveo U200 accelerator card),...
-
A self-optimization mechanism for generalized adaptive notch smoother
PublicationTracking of nonstationary narrowband signals is often accomplished using algorithms called adaptive notch filters (ANFs). Generalized adaptive notch smoothers (GANSs) extend the concepts of adaptive notch filtering in two directions. Firstly, they are designed to estimate coefficients of nonstationary quasi-periodic systems, rather than signals. Secondly, they employ noncausal processing, which greatly improves their accuracy and...
-
All-gather Algorithms Resilient to Imbalanced Process Arrival Patterns
PublicationTwo novel algorithms for the all-gather operation resilient to imbalanced process arrival patterns (PATs) are presented. The first one, Background Disseminated Ring (BDR), is based on the regular parallel ring algorithm often supplied in MPI implementations and exploits an auxiliary background thread for early data exchange from faster processes to accelerate the performed all-gather operation. The other algorithm, Background Sorted...
-
Acceleration of Electromagnetic Simulations on Reconfigurable FPGA Card
PublicationIn this contribution, the hardware acceleration of electromagnetic simulations on the reconfigurable field-programmable-gate-array (FPGA) card is presented. In the developed implementation of scientific computations, the matrix-assembly phase of the method of moments (MoM) is accelerated on the Xilinx Alveo U200 card. The computational method involves discretization of the frequency-domain mixed potential integral equation using...
-
Mathematical modelling of two-step nitrification-denitrification for treatment of sludge digester liquors: influence of nitrite (NO2-N) on the process kinetics
PublicationSeparate treatment of the sludge digester liquors is an alternative for expansion of the mainstream treatment line. In order to reduce the oxygen demand for nitrification and organic carbon demand for denitrification, a shortcut in the nitrogen conversion pathway has been promoted in recent years, i.e. nitrification-denitrification via NO2-N instead of NO3-N. Although NO2-N is a common intermediate product of nitrification and...
-
Application of mechanistic and data-driven models for nitrogen removal in wastewater treatment systems
PublicationIn this dissertation, the application of mechanistic and data-driven models in nitrogen removal systems including nitrification and deammonification processes was evaluated. In particular, the influential parameters on the activity of the Nitrospira activity were assessed using response surface methodology (RSM). Various long-term biomass washout experiments were operated in two parallel sequencing batch reactor (SBR) with a different...
-
Object serialization and remote exception pattern for distributed C++/MPI application
PublicationMPI is commonly used standard in development of scientific applications. It focuses on interlanguage operability and is not very well object oriented. The paper proposes a general pattern enabling design of distributed and object oriented applications. It also presents its sample implementations and performance tests.
-
Towards Easy-to-Use Checkpointing of MPI Applications within CLUSTERIX.
PublicationW literaturze wymienia się wiele bibliotek/systemów zarówno poziomu jądra jak i użytkownika, które wspomagają zapisywanie i odtwarzanie stanu procesów. W odniesieniu do aplikacji równoległych, jest to jednak zadanie cały czas trudne. Praca prezentuje nasze podejście do zapisywania/odtwarzania stanu aplikacji MPI wspomagane przez programistę, które wykorzystane będzie w środowisku projektu CLUSTERIX tj. zintegrowanej grupie klastrów...
-
An extension of the method of quasilinearization
PublicationMetodę kwazilinearyzacji zastosowano do problemów początkowych gdy prawą stronę zagadnienia można przedstawić za pomocą nieliniowej funkcji "rozszerzenia", zakładając o niej pewną regularność. Pokazano, że odpowiednio skonstruowane ciągi monotoniczne są zbieżne kwadratowo do rozwiązania problemu. Praca uogólnia odpowiednie wyniki, gdy prawa strona jest sumą funkcji wklęsłych i wypukłych ze względu na ostatni argument.
-
JOURNAL OF EXTENSION
Journals -
Modeling Parallel Applications in the MERPSYS Environment
PublicationThe chapter presents how to model parallel computational applications for which simulation of execution in a large-scale parallel or distributed environment is performed within the MERPSYS environment. Specifically, it is shown what approaches can be adopted to model key paradigms often used for parallel applications: master-slave, geometric parallelism (single program multiple data), pipelined and divide-and-conquer applications....
-
An extension to the FEEDB Multimodal Database of Facial Expressions and Emotions
PublicationFEEDB is a multimodal database that contains recordings of people expressing different emotions, captured by using a Microsoft Kinect sensor. Data were originally provided in the device’s proprietary format (XED), requiring both the Microsoft Kinect Studio application and a Kinect sensor attached to the system to use the files. In this paper, we present an extension of the database. For a selection of recordings, we also provide...
-
Efektywna warstwa pośrednicząca dla obliczeń typu master-slave w środowisku C++/MPI
PublicationPokazano, jak dla wysokowydajnościowego algorytmu pisanego w modelu master-slave w języku C++ i spełniającego pewne ograniczenia można napisać i wykorzystać warstwę komunikacyjną zupełnie oddzielającą kod odpowiedzialny za komunikację od kodu odpowiedzialnego za dzie-dzinę problemową. Przedstawiona zostaje specyfkacja wymagań, jakie powinien spełniać hipotetyczny system rozproszony oraz warstwa komunikacyjna, a także wymagania...
-
Three levels of fail-safe mode in MPI I/O NVRAM distributed cache
PublicationThe paper presents architecture and design of three versions for fail-safe data storage in a distributed cache using NVRAM in cluster nodes. In the first one, cache consistency is assured through additional buffering write requests. The second one is based on additional write log managers running on different nodes. The third one benefits from synchronization with a Parallel File System (PFS) for saving data into a new file which...
-
The evaluation of eGlasses eye tracking module as an extension for Scratch
PublicationIn this paper we present the possibility of using eGlasses eye tracking module as an extension for Scratch programming tool which is a visual programming language supporting computer skills learning. The main concept behind this project is to setup the interface for rapid interaction design. Eye tracking is a powerful tool for hands free communication but for that requires a dedicated software. This software is rarely tailored...
-
Parallel immune system for graph coloring
PublicationThis paper presents a parallel artificial immune system designed forgraph coloring. The algorithm is based on the clonal selection principle. Each processor operates on its own pool of antibodies and amigration mechanism is used to allow processors to exchange information. Experimental results show that migration improves the performance of the algorithm. The experiments were performed using a high performance cluster on a set...
-
Parallel Computations of Text Similarities for Categorization Task
PublicationIn this chapter we describe the approach to parallel implementation of similarities in high dimensional spaces. The similarities computation have been used for textual data categorization. A test datasets we create from Wikipedia articles that with their hyper references formed a graph used in our experiments. The similarities based on Euclidean distance and Cosine measure have been used to process the data using k-means algorithm....