Computational steering is a new alternative approach to classical high performance computing studies. In contrast to the traditional pipeline of separated preprocessing, computation and postprocessing steps computatio...
详细信息
ISBN:
(纸本)9781905088423
Computational steering is a new alternative approach to classical high performance computing studies. In contrast to the traditional pipeline of separated preprocessing, computation and postprocessing steps computational steering joins these into one simultaneous work flow. This paper will describe a computational steering environment enabling its user to visualize simulation results on-the-fly and to interact with the running simulation. The main modules constituting the computational steering application, i.e. the simulation kernel, preprocessing, postprocessing and steering, as well as the communication layout between them will be introduced. The main focus will be laid on the optimization of the data flow between these components.
This paper shows an efficient algorithm to resolve compressible reacting flow with Direct Numerical Simulation (DNS) method. It has been widely known that DNS of a reacting jet flow in the laboratory scale is a huge c...
详细信息
ISBN:
(纸本)9781905088423
This paper shows an efficient algorithm to resolve compressible reacting flow with Direct Numerical Simulation (DNS) method. It has been widely known that DNS of a reacting jet flow in the laboratory scale is a huge challenge and of vital importance. This article first present a test to identify the effectivity of the parallel algorithm preparing for further large-scale computation and then design a grid system suited to an experimental turbulent lifted jet flame to perform DNS available. Finally, a comparison between the DNS results and measured data will be shown to verify whether we can resolve this experimental flame well.
Since its early heuristic development, direct search methods (DSM) have proved to be robust and reliable, both from theoretical and practical perspectives. Some of these algorithms are now able to solve noisy, nonsmoo...
详细信息
ISBN:
(纸本)9781905088423
Since its early heuristic development, direct search methods (DSM) have proved to be robust and reliable, both from theoretical and practical perspectives. Some of these algorithms are now able to solve noisy, nonsmooth or nonconvex problems and its algorithmic structure favours parallelization, drawing the attention of the structural optimization community as a promising alternative to the use of meta-heuristics. In the present work, the performance of DSM is compared to genetic algorithms, when solving hard or expensive structural optimization problems. parallel implementations are considered when large computational times are involved in function evaluation.
One of the key challenges of IT for architecture, engineering and construction is to provide the efficient, effective and flexible access to information and to provide all possible channels of communication. The world...
详细信息
ISBN:
(纸本)9781905088423
One of the key challenges of IT for architecture, engineering and construction is to provide the efficient, effective and flexible access to information and to provide all possible channels of communication. The world today is becoming increasingly mobile and users often do their work with the help of several different devices during their day. That is why the unified access to networks, servers, storage, applications, and services is needed. In order to achieve that mentioned resources cannot reside on local devices anymore, they should rather live in "the cloud" and be accessible via Web services to a vast spectrum of devices from different locations. cloudcomputing is still an evolving paradigm that is rather difficult to define but it usually described as a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources that can be rapidly provisioned and released with minimal management effort or service provider interaction. The continuous process improvement via the strategic implementation of innovative information and communication technologies is essential for the long-term survival of construction firms. Unfortunately because of the specifics of the AEC sector SME construction companies have difficulties keeping up with the pace of the advancements in the technology and software. The paper examines the cloudcomputing model and other upcoming technologies and trends through civil engineering SME specific requirements.
Multiparticle Collision Dynamics (MPC) is a technique to simulate the hydrodynamic properties of fluid and flow phenomena at the mesoscale. It is a particle-based approach which bases on a free streaming of the partic...
详细信息
ISBN:
(纸本)9781905088423
Multiparticle Collision Dynamics (MPC) is a technique to simulate the hydrodynamic properties of fluid and flow phenomena at the mesoscale. It is a particle-based approach which bases on a free streaming of the particles and a stochastic rotation of the particle velocities. To accelerate the MPC algorithm, the Cell Broadband Engine (Cell/BE) is considered as target platform for a parallel implementation. For portability reasons the programming model Cell Superscalar (CellSs) is used which addresses the automatic exploitation of functional parallelism from a sequential application with annotations. To develop an implementation with a high level of parallelism, the MPC algorithm is redesigned to a domain decomposition approach which consideres the Cell/BE and CellSs specific requirements.
We describe a computational framework for coupling an array of parallel numerical solvers, each designed to tackle a flow problem at different spatial and temporal scale. Specifically, we describe an architecture of t...
详细信息
ISBN:
(纸本)9781905088423
We describe a computational framework for coupling an array of parallel numerical solvers, each designed to tackle a flow problem at different spatial and temporal scale. Specifically, we describe an architecture of the ultraparallel metasolver N epsilon kT alpha rG for multiscale blood flow simulations. We focus on the multi-level communicating interface (MCI) which provides an efficient communication between hundreds of thousands of computer processors used by the solvers. Performance is evaluated using specifically designed benchmark problems and also a "real-life" simulation of a blood flow in the arterial network of the brain.
This paper describes application of domain decomposition methods to coupled heat and moisture transfer problems. The finite element method is used for space discretization while the trapezoidal method is used for time...
详细信息
ISBN:
(纸本)9781905088423
This paper describes application of domain decomposition methods to coupled heat and moisture transfer problems. The finite element method is used for space discretization while the trapezoidal method is used for time discretization. Many material models for coupled problems lead to nonsymmetric systems of algebraic equations and therefore the Schur complement method is considered. In this paper, the Kunzel model is employed. Application of the Schur complement method is shown on an example of analysis of heat and moisture transfer in building brick. Decomposition of the original mesh into four and eight subdomains reduces computational time significantly and very good speedup is obtained.
We present a workflow-based algorithm for identifying threads to an urban water management system. Through gridcomputing we provide the necessary high-performance computing resources to deliver quickly solutions to t...
详细信息
ISBN:
(纸本)9781905088423
We present a workflow-based algorithm for identifying threads to an urban water management system. Through gridcomputing we provide the necessary high-performance computing resources to deliver quickly solutions to the problem. We prototyped a new middleware called cyberaide, that enables easy access to grid resources through portals or the command line. A workflow system is used to manage resources in fault tolerant fashion. In addition, we contrast the architecture with a Hadoop implementation. Resources from Teragrid and Futuregrid are used to test the feasibility of using the toolkit for a scientific application.
An algorithm for uniform PRN generation on parallel computers, providing concordant formation of the segments of a random numbers sequence on each processor, is considered. A class of rarified polynomials is determine...
详细信息
ISBN:
(纸本)9781905088423
An algorithm for uniform PRN generation on parallel computers, providing concordant formation of the segments of a random numbers sequence on each processor, is considered. A class of rarified polynomials is determined which allows the reduction of the numerical complexity and to increase the level of inherent parallelism of PRN generation algorithms. The high quality of generated sequences is confirmed by Diehard battery of tests and by percolation threshold determination. The library for parallel PRN generation on NVIDIA GPUs is created. The performance of generators developed for CPU and GPU was compared with the performance of the available libraries of generators. The results are presented which confirm the efficiency of the developed methods.
In this paper, we describe the challenges faced and lessons learned while establishing a large-scale high performance cloudcomputing service that enables online mechanical structural analysis and many other scientifi...
详细信息
ISBN:
(纸本)9781905088423
In this paper, we describe the challenges faced and lessons learned while establishing a large-scale high performance cloudcomputing service that enables online mechanical structural analysis and many other scientific applications using the finite element analysis (FEA) technique. The service is intended to process many independent and loosely-dependent (e.g. assembled system) tasks concurrently. Challenges faced include accurate job characterization, handling of many-task mixed jobs, sensitivity of task execution to multi-threading parameters, effective multi-core scheduling in a single node, and achieving seamless scale across multiple nodes. We find that significant performance gains in terms of both job completion latency and throughput are possible via dynamic or "smart" partitioning and resource-aware scheduling compared to shortest first and aggressive job scheduling techniques. We also discuss issues related to secure and private processing of sensitive models in the cloud.
暂无评论