GRASS GIS ON HIGH PERFORMANCE COMPUTING WITH MPI, OPENMP AND NINF-G PROGRAMMING FRAMEWORK
GRASS GIS ON HIGH PERFORMANCE COMPUTING WITH MPI, OPENMP AND NINF-G PROGRAMMING FRAMEWORK
复制标题
DOI：
--
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
S. Akhter;K. Aida;Y. Chemin
中科院分区：
文献类型：
--
作者：
S. Akhter;K. Aida;Y. Chemin
GRASS GIS（地理资源分析支持系统）是一个免费的开源软件，用于遥感（RS）和地理信息系统（GIS）数据分析和可视化。在GRASS内部，已经开发了不同的模块来处理卫星图像。目前，GRASS使用数据库来处理大型数据集，通过将GRASS模块与并行和分布式计算相结合，可以大大提高GRASS处理大型数据集的性能和能力。基于多台计算机的分布式系统（集群和网格）以较低的成本具有较大的处理能力，自然，选择转向开发高性能计算（HPC）应用程序。然而，将GRASS模块直接移植到HPC环境并不是一件容易的事。卫星图像处理应用程序的开发人员需要解决数据和任务的分布问题，或者如何在单个或多个集群环境中分布数据和任务。高性能计算的工作量、带宽、处理器速度、评估方法的参数和数据量都是额外的考虑因素。GRASS模块，即i)“i”“vi”是由Kamble和Chemin（2006）开发的，用于处理13个植被指数。“lmf”是由Akhter等人（2008）开发的，用于从RS图像中去除大气影响。“gaswap”是由Akhter等人（2006）开发的，用于找出从RS图像中无法直接看到的作物参数，将作为在HPC上开发GRASS模块框架的三个案例研究进行讨论。开发能够在高性能计算系统上运行GRASS GIS环境的RS图像处理方法，将是本文研究的主要问题。此外，本文还将讨论分布式GRASS模型在三种不同编程平台（MPI、nf- g和OpenMP）上的不同实现，并介绍它们的性能。
GRASS GIS (Geographical Resources Analysis Support System) is a free, open source software and has been used for Remote Sensing (RS) and Geographic Information System (GIS) data analysis and visualization. Inside GRASS, different modules have been developed for processing satellite images. Currently, GRASS uses databases to handle large datasets and the performance and capabilities of GRASS for large datasets can be greatly improved by integrating GRASS modules with parallel and distributed computing. Multi computer based distributed systems (clusters and Grids) have a large processing capacity for a lower cost, naturally, choice turns towards developing High Performance Computing (HPC) applications. However, it is not an easy job to port GRASS modules directly to HPC environment. The developers of satellite image processing applications need to solve the problem of both data and task distribution, or how to distribute data and tasks among single or multiple clusters environment. The workload in HPC, the bandwidth, the processors speed, parameters of evaluation methods and data size are additional concerning factors. GRASS modules, i.e. i) “i.vi” is developed by Kamble and Chemin (2006) to process 13 vegetation indices, ii) “i.lmf” is developed by Akhter et al. (2008) to remove the atmospheric effects from RS images and iii) “r.gaswap” is developed by Akhter et al. (2006) to find out the crop parameters those are not directly visible from RS images, will be discussed as three case studies to developed GRASS module framework on HPC. Developing the methodology, which enables to run GRASS GIS environment for RS images processing on HPC systems, will be the main concerning issue of this paper. Additionally, different implementations for distributed GRASS models will be discussed on three different programming platforms (MPI, Ninf-G and OpenMP) and their performance will also be presented in this paper.