Assessing improvements to the parallel volume rendering pipeline at large scale

Tom Peterka, Robert Ross, Hongfeng Yu, Kwan-Liu Ma, Wesley Kendall, Jian Huang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)

Abstract

Computational science's march toward the petascale demands innovations in analysis and visualization of the resulting datasets. As scientists generate terabyte and petabyte data, it is insufficient to measure the performance of visual analysis algorithms by rendering speed only, because performance is dominated by data movement. We take a systemwide view in analyzing the performance of software volume rendering on the IBM Blue Gene/P at over 10,000 cores by examining the relative costs of the I/O, rendering, and compositing portions of the volume rendering algorithm. This examination uncovers room for improvement in data input, load balancing, memory usage, image compositing, and image output. We present four improvements to the basic algorithm to address these bottlenecks. We show the benefit of an alternative rendering distribution scheme that improves load balance, and how to scale memory usage so that large data and image sizes do not overload system memory. To improve compositing, we experiment with a hybrid MPI-multithread programming model, and to mitigate the high cost of I/O, we implement multiple parallel pipelines to partially hide the I/O cost when rendering many time steps. Measuring the benefits of these techniques at scale reinforces the conclusion that BG/P is an effective platform for volume rendering of large datasets and that our volume rendering algorithm, enhanced by the techniques presented here, scales to large problem and system sizes.

Original languageEnglish (US)
Title of host publication2008 Workshop on Ultrascale Visualization, UltraVis 2008
Pages13-23
Number of pages11
DOIs
StatePublished - Dec 1 2008
Event2008 Workshop on Ultrascale Visualization, UltraVis 2008 - Austin, TX, United States
Duration: Nov 16 2008Nov 16 2008

Other

Other2008 Workshop on Ultrascale Visualization, UltraVis 2008
CountryUnited States
CityAustin, TX
Period11/16/0811/16/08

Fingerprint

Volume rendering
Pipelines
Data storage equipment
Costs
Resource allocation
Visualization
Genes
Innovation
Experiments

Keywords

  • Distributed scientific visualization
  • Load balancing
  • Multithreading
  • Parallel volume rendering
  • Scalability
  • Software raycasting

ASJC Scopus subject areas

  • Computer Science Applications
  • Computer Vision and Pattern Recognition

Cite this

Peterka, T., Ross, R., Yu, H., Ma, K-L., Kendall, W., & Huang, J. (2008). Assessing improvements to the parallel volume rendering pipeline at large scale. In 2008 Workshop on Ultrascale Visualization, UltraVis 2008 (pp. 13-23). [5154059] https://doi.org/10.1109/ULTRAVIS.2008.5154059

Assessing improvements to the parallel volume rendering pipeline at large scale. / Peterka, Tom; Ross, Robert; Yu, Hongfeng; Ma, Kwan-Liu; Kendall, Wesley; Huang, Jian.

2008 Workshop on Ultrascale Visualization, UltraVis 2008. 2008. p. 13-23 5154059.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Peterka, T, Ross, R, Yu, H, Ma, K-L, Kendall, W & Huang, J 2008, Assessing improvements to the parallel volume rendering pipeline at large scale. in 2008 Workshop on Ultrascale Visualization, UltraVis 2008., 5154059, pp. 13-23, 2008 Workshop on Ultrascale Visualization, UltraVis 2008, Austin, TX, United States, 11/16/08. https://doi.org/10.1109/ULTRAVIS.2008.5154059
Peterka T, Ross R, Yu H, Ma K-L, Kendall W, Huang J. Assessing improvements to the parallel volume rendering pipeline at large scale. In 2008 Workshop on Ultrascale Visualization, UltraVis 2008. 2008. p. 13-23. 5154059 https://doi.org/10.1109/ULTRAVIS.2008.5154059
Peterka, Tom ; Ross, Robert ; Yu, Hongfeng ; Ma, Kwan-Liu ; Kendall, Wesley ; Huang, Jian. / Assessing improvements to the parallel volume rendering pipeline at large scale. 2008 Workshop on Ultrascale Visualization, UltraVis 2008. 2008. pp. 13-23
@inproceedings{ea47d431cb9542dd9eb1a3774482f505,
title = "Assessing improvements to the parallel volume rendering pipeline at large scale",
abstract = "Computational science's march toward the petascale demands innovations in analysis and visualization of the resulting datasets. As scientists generate terabyte and petabyte data, it is insufficient to measure the performance of visual analysis algorithms by rendering speed only, because performance is dominated by data movement. We take a systemwide view in analyzing the performance of software volume rendering on the IBM Blue Gene/P at over 10,000 cores by examining the relative costs of the I/O, rendering, and compositing portions of the volume rendering algorithm. This examination uncovers room for improvement in data input, load balancing, memory usage, image compositing, and image output. We present four improvements to the basic algorithm to address these bottlenecks. We show the benefit of an alternative rendering distribution scheme that improves load balance, and how to scale memory usage so that large data and image sizes do not overload system memory. To improve compositing, we experiment with a hybrid MPI-multithread programming model, and to mitigate the high cost of I/O, we implement multiple parallel pipelines to partially hide the I/O cost when rendering many time steps. Measuring the benefits of these techniques at scale reinforces the conclusion that BG/P is an effective platform for volume rendering of large datasets and that our volume rendering algorithm, enhanced by the techniques presented here, scales to large problem and system sizes.",
keywords = "Distributed scientific visualization, Load balancing, Multithreading, Parallel volume rendering, Scalability, Software raycasting",
author = "Tom Peterka and Robert Ross and Hongfeng Yu and Kwan-Liu Ma and Wesley Kendall and Jian Huang",
year = "2008",
month = "12",
day = "1",
doi = "10.1109/ULTRAVIS.2008.5154059",
language = "English (US)",
isbn = "9781424428618",
pages = "13--23",
booktitle = "2008 Workshop on Ultrascale Visualization, UltraVis 2008",

}

TY - GEN

T1 - Assessing improvements to the parallel volume rendering pipeline at large scale

AU - Peterka, Tom

AU - Ross, Robert

AU - Yu, Hongfeng

AU - Ma, Kwan-Liu

AU - Kendall, Wesley

AU - Huang, Jian

PY - 2008/12/1

Y1 - 2008/12/1

N2 - Computational science's march toward the petascale demands innovations in analysis and visualization of the resulting datasets. As scientists generate terabyte and petabyte data, it is insufficient to measure the performance of visual analysis algorithms by rendering speed only, because performance is dominated by data movement. We take a systemwide view in analyzing the performance of software volume rendering on the IBM Blue Gene/P at over 10,000 cores by examining the relative costs of the I/O, rendering, and compositing portions of the volume rendering algorithm. This examination uncovers room for improvement in data input, load balancing, memory usage, image compositing, and image output. We present four improvements to the basic algorithm to address these bottlenecks. We show the benefit of an alternative rendering distribution scheme that improves load balance, and how to scale memory usage so that large data and image sizes do not overload system memory. To improve compositing, we experiment with a hybrid MPI-multithread programming model, and to mitigate the high cost of I/O, we implement multiple parallel pipelines to partially hide the I/O cost when rendering many time steps. Measuring the benefits of these techniques at scale reinforces the conclusion that BG/P is an effective platform for volume rendering of large datasets and that our volume rendering algorithm, enhanced by the techniques presented here, scales to large problem and system sizes.

AB - Computational science's march toward the petascale demands innovations in analysis and visualization of the resulting datasets. As scientists generate terabyte and petabyte data, it is insufficient to measure the performance of visual analysis algorithms by rendering speed only, because performance is dominated by data movement. We take a systemwide view in analyzing the performance of software volume rendering on the IBM Blue Gene/P at over 10,000 cores by examining the relative costs of the I/O, rendering, and compositing portions of the volume rendering algorithm. This examination uncovers room for improvement in data input, load balancing, memory usage, image compositing, and image output. We present four improvements to the basic algorithm to address these bottlenecks. We show the benefit of an alternative rendering distribution scheme that improves load balance, and how to scale memory usage so that large data and image sizes do not overload system memory. To improve compositing, we experiment with a hybrid MPI-multithread programming model, and to mitigate the high cost of I/O, we implement multiple parallel pipelines to partially hide the I/O cost when rendering many time steps. Measuring the benefits of these techniques at scale reinforces the conclusion that BG/P is an effective platform for volume rendering of large datasets and that our volume rendering algorithm, enhanced by the techniques presented here, scales to large problem and system sizes.

KW - Distributed scientific visualization

KW - Load balancing

KW - Multithreading

KW - Parallel volume rendering

KW - Scalability

KW - Software raycasting

UR - http://www.scopus.com/inward/record.url?scp=70350766689&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70350766689&partnerID=8YFLogxK

U2 - 10.1109/ULTRAVIS.2008.5154059

DO - 10.1109/ULTRAVIS.2008.5154059

M3 - Conference contribution

AN - SCOPUS:70350766689

SN - 9781424428618

SP - 13

EP - 23

BT - 2008 Workshop on Ultrascale Visualization, UltraVis 2008

ER -