高性能并行可视化服务器的资源管理技术研究
Research on the Resources Management Technique of High-Performance Parallel Visualization Server

作者: 路 石 , 孟创斌 , 李思昆 , 王文珂 , 曾 亮 :国防科学技术大学高性能计算国家重点实验室,长沙;

关键词: 可视化服务器任务调度资源管理高性能并行计算Visualization Server Task Scheduling Resources Management High-Performance Parallel Compute

摘要:
开发高性能并行可视化服务器可充分发挥高性能计算机的资源优势,实现基于高性能计算机的高效并行可视化服务,克服传统后处理模式科学可视化存在的效率低等问题。本文介绍了高性能并行可视化服务器的功能和组成结构,重点论述了在研究高性能并行可视化服务器资源管理技术中提出的基于任务属性的计算结点资源分配算法和基于线性回归的任务属性自适应维护算法,算法能够有效利用高性能计算机的计算资源完成科学计算可视化应用任务的计算节点分配,并具有良好的任务属性自适应维护功能。实验结果表明所提出的算法可针对大数据科学计算可视化任务特点,有效完成并行可视计算的任务调度和资源分配,提高科学计算可视化的效率。

Abstract: Developing high-performance parallel visualization server can give full play to the advantages of resources in high-performance computer, provide efficient parallel visualization service based on high-performance computer, and overcome the low efficiency of after-treatment model in the tra-ditional visualization way. This thesis introduces the structure and functions of the high-perfor- mance parallel visualization server; focuses on the algorithm of resource allocation and optimizing in the high-performance parallel visualization server. The algorithm can allocate compute nodes with the resources of the high-performance computer for the scientific computation visualization application efficiently, and have a strong self-adapted ability. The result of the experiment indicates that our algorithm completed the allocation of computing resources efficiently according to the characteristics of big data scientific computation visualization tasks, and improved the efficiency of scientific computation visualization very much compared to the traditional visualization model.

文章引用: 路 石 , 孟创斌 , 李思昆 , 王文珂 , 曾 亮 (2014) 高性能并行可视化服务器的资源管理技术研究。 软件工程与应用, 3, 131-143. doi: 10.12677/SEA.2014.35016

参考文献

[1] Yu, H., Tu, T., Bielak, J., Ghattas, O., Lopez, J.C., Ma, K.-L., O’Hallaron, D.R., Ramirez-Guzman, L., Stone, N., Ta-borda-Rios, R. and Urbanic, J. (2006) Remote runtime steering of integrated terascale simulation and visualization, 2006. HPC Analytics Challenge, ACM/IEEE Supercomputing 2006 Conference. http://www.cs.cmu.edu/~quake/quakeshow.pdf

[2] ParaView catalyst for in situ analysis. http://www.paraview.org/in-situ/

[3] Visit graphical user interface help window. https://wci.llnl.gov/simulation/computer-codes/visit/screenshots4

[4] Biddiscombe, J., Soumagne, J., Ogre, G., et al. (2011) Parallel computational steering and analysis for HPC applications using a paraview interface and the HDF5 DSM virtural file driver. Proceedings of the 11th Eurographics Conference on Parallel Graphics and Visualization, 91-100.

[5] 王观玉 (2011) 网格计算中任务调度算法的研究个改进. 计算机工程与科学, 10, 186-190.

[6] 陈亮 (2013) 基于并行资源属性选择的任务调度系统模型的研究. 硕士论文, 电子科技大学, 成都.

[7] 沈瑜, 李娟, 常飚, 孙静 (2014) 高性能计算机统一资源管理系统的设计与实现. 计算技术与自动化, 1, 83-90.

[8] SLURM (2011) A highly scalable resource manager.

[9] IBM Corporation (2004) IBM Loadleveler for AIX 5L and Linx. Using and Administering. IBM Corporation, New York.

[10] 张洋, 陈文波, 李廉 (2007) 高性能集群作业管理系统TORQUE分析与应用实现. 计算机工程与科学, 10, 132-134.

[11] 吴孟达, 李兵, 汪文浩 (2004) 高等工程数学. 科学出版社, 北京.

分享
Top