Zhengchun Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Zhengchun Liu has a total of 117 co-authors. Here are the top co-authors by number of shared publications:
Rajkumar Kettimuthu
Data Science and Learning Division Argonne National Laboratory 60439 Lemont IL USA
21 shared publications- Ian Foster
Argonne National Laboratory, Argonne, IL, USA
17 shared publications Ian Foster
Data Science and Learning Division Argonne National Laboratory 60439 Lemont IL USA
13 shared publications- Rajkumar Kettimuthu
Argonne National Laboratory
9 shared publications - Dolores Rexachs
Computer Architecture and Operating Systems Department, Universitat Autònoma de Barcelona, Bellaterra, Barcelona, Spain
6 shared publications - Veerabhadra Rao Kotamarthi
Argonne National Laboratory, 9700 Cass Ave., Lemont, IL, 60439
6 shared publications
Latest Publications
- Open Access
- 17 July 2024
- Published by International Union of Crystallography (IUCr) in Journal of Applied Crystallography
- Vol. 57 (4) , 1158-1170
Rapid detection of rare events from in situ X-ray diffraction data using machine learning
High-energy X-ray diffraction methods can non-destructively map the 3D microstructure and associated attributes of metallic polycrystalline engineering materials in their bulk form. These methods are often combined with external stimuli such as thermo-mechanical loading to take snapshots of the evolving microstructure and attributes over time. However, the extreme data volumes and the high costs of traditional data acquisition and reduction approaches pose a barrier to quickly extracting actionable insights and improving the temporal resolution of these snapshots. This article presents a fully automated technique capable of rapidly detecting the onset of plasticity in high-energy X-ray microscopy data. The technique is computationally faster by at least 50 times than the traditional approaches and works for data sets that are up to nine times sparser than a full data set. This new technique leverages self-supervised image representation learning and clustering to transform massive data sets into compact, semantic-rich representations of visually salient characteristics ( e.g. peak shapes). These characteristics can rapidly indicate anomalous events, such as changes in diffraction peak shapes. It is anticipated that this technique will provide just-in-time actionable information to drive smarter experiments that effectively deploy multi-modal X-ray diffraction methods spanning many decades of length scales.
- 17 July 2024
- Published by International Union of Crystallography (IUCr) in Journal of Applied Crystallography
- Vol. 57 (4) , 1158-1170
Open AccessRapid detection of rare events from in situ X-ray diffraction data using machine learning
High-energy X-ray diffraction methods can non-destructively map the 3D microstructure and associated attributes of metallic polycrystalline engineering materials in their bulk form. These methods are often combined with external stimuli such as thermo-mechanical loading to take snapshots of the evolving microstructure and attributes over time. However, the extreme data volumes and the high costs of traditional data acquisition and reduction approaches pose a barrier to quickly extracting actionable insights and improving the temporal resolution of these snapshots. This article presents a fully automated technique capable of rapidly detecting the onset of plasticity in high-energy X-ray microscopy data. The technique is computationally faster by at least 50 times than the traditional approaches and works for data sets that are up to nine times sparser than a full data set. This new technique leverages self-supervised image representation learning and clustering to transform massive data sets into compact, semantic-rich representations of visually salient characteristics ( e.g. peak shapes). These characteristics can rapidly indicate anomalous events, such as changes in diffraction peak shapes. It is anticipated that this technique will provide just-in-time actionable information to drive smarter experiments that effectively deploy multi-modal X-ray diffraction methods spanning many decades of length scales.
- Open Access
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 269-279
Intelligent Scaling in Amazon Redshift
Cloud-based data warehouses are built to be easy to use, requiring minimal intervention from customers as their workloads scale. However, there are still many dimensions of a workload that they do not scale with automatically. For example, in cloud-managed clusters, large ad-hoc queries and ETL workloads must use the same cluster size provisioned for the rest of the workload, and warehouse size does not automatically grow as the underlying data grows in size, causing queries to slow down. In this paper, we describe RAIS, the latest collection of AI-powered scaling and optimization techniques in Amazon Redshift, released in preview at re:Invent 2023, which enable it to scale both vertically and horizontally to adapt to all types of workload variability. RAIS dynamically provisions compute resources to run heavy queries efficiently and automatically optimizes warehouse size for the customer's workload, even as it shifts over time. We show that, depending on the workload, RAIS improves either cost or average query execution time by up to 7.6× and 14.2×, respectively, over existing baselines.
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 269-279
Open AccessIntelligent Scaling in Amazon Redshift
Cloud-based data warehouses are built to be easy to use, requiring minimal intervention from customers as their workloads scale. However, there are still many dimensions of a workload that they do not scale with automatically. For example, in cloud-managed clusters, large ad-hoc queries and ETL workloads must use the same cluster size provisioned for the rest of the workload, and warehouse size does not automatically grow as the underlying data grows in size, causing queries to slow down. In this paper, we describe RAIS, the latest collection of AI-powered scaling and optimization techniques in Amazon Redshift, released in preview at re:Invent 2023, which enable it to scale both vertically and horizontally to adapt to all types of workload variability. RAIS dynamically provisions compute resources to run heavy queries efficiently and automatically optimizes warehouse size for the customer's workload, even as it shifts over time. We show that, depending on the workload, RAIS improves either cost or average query execution time by up to 7.6× and 14.2×, respectively, over existing baselines.
- Open Access
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 280-294
Stage: Query Execution Time Prediction in Amazon Redshift
Query performance (e.g., execution time) prediction is a critical component of modern DBMSes. As a pioneering cloud data warehouse, Amazon Redshift relies on an accurate execution time prediction for many downstream tasks, ranging from high-level optimizations, such as automatically creating materialized views, to low-level tasks on the critical path of query execution, such as admission, scheduling, and execution resource control. Unfortunately, many existing execution time prediction techniques, including those used in Redshift, suffer from cold start issues, inaccurate estimation, and are not robust against workload/data changes. In this paper, we propose a novel hierarchical execution time predictor: the Stage predictor. The Stage predictor is designed to leverage the unique characteristics and challenges faced by Redshift. The Stage predictor consists of three model states: an execution time cache, a lightweight local model optimized for a specific DB instance with uncertainty measurement, and a complex global model that is transferable across all instances in Redshift. We design a systematic approach to use these models that best leverages optimality (cache), instance-optimization (local model), and transferable knowledge about Redshift (global model). Experimentally, we show that the Stage predictor makes more accurate and robust predictions while maintaining a practical inference latency and memory overhead. Overall, the Stage predictor can improve the average query execution latency by 20% on these instances compared to the prior query performance predictor in Redshift.
- 9 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 280-294
Open AccessStage: Query Execution Time Prediction in Amazon Redshift
Query performance (e.g., execution time) prediction is a critical component of modern DBMSes. As a pioneering cloud data warehouse, Amazon Redshift relies on an accurate execution time prediction for many downstream tasks, ranging from high-level optimizations, such as automatically creating materialized views, to low-level tasks on the critical path of query execution, such as admission, scheduling, and execution resource control. Unfortunately, many existing execution time prediction techniques, including those used in Redshift, suffer from cold start issues, inaccurate estimation, and are not robust against workload/data changes. In this paper, we propose a novel hierarchical execution time predictor: the Stage predictor. The Stage predictor is designed to leverage the unique characteristics and challenges faced by Redshift. The Stage predictor consists of three model states: an execution time cache, a lightweight local model optimized for a specific DB instance with uncertainty measurement, and a complex global model that is transferable across all instances in Redshift. We design a systematic approach to use these models that best leverages optimality (cache), instance-optimization (local model), and transferable knowledge about Redshift (global model). Experimentally, we show that the Stage predictor makes more accurate and robust predictions while maintaining a practical inference latency and memory overhead. Overall, the Stage predictor can improve the average query execution latency by 20% on these instances compared to the prior query performance predictor in Redshift.
- Open Access
- 7 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 190-200
MalleTrain: Deep Neural Networks Training on Unfillable Supercomputer Nodes
First-come first-serve scheduling can result in substantial (up to 10%) of transiently idle nodes on supercomputers. Recognizing that such unfilled nodes are well-suited for deep neural network (DNN) training, due to the flexible nature of DNN training tasks, Liu et al. proposed that the re-scaling DNN training tasks to fit gaps in schedules be formulated as a mixed-integer linear programming (MILP) problem, and demonstrated via simulation the potential benefits of the approach. Here, we introduce MalleTrain, a system that provides the first practical implementation of this approach and that furthermore generalizes it by allowing it to be used even for DNN training applications for which model information is unknown before runtime. Key to this latter innovation is the use of a lightweight online job profiling advisor (JPA) to collect critical scalability information for DNN jobs---information that it then employs to optimize resource allocations dynamically, in real time. We describe the MalleTrain architecture and present the results of a detailed experimental evaluation on a supercomputer GPU cluster and several representative DNN training workloads, including neural architecture search and hyperparameter optimization. Our results not only confirm the practical feasibility of leveraging idle supercomputer nodes for DNN training but improve significantly on prior results, improving training throughput by up to 22.3% without requiring users to provide job scalability information.
- 7 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 190-200
Open AccessMalleTrain: Deep Neural Networks Training on Unfillable Supercomputer Nodes
First-come first-serve scheduling can result in substantial (up to 10%) of transiently idle nodes on supercomputers. Recognizing that such unfilled nodes are well-suited for deep neural network (DNN) training, due to the flexible nature of DNN training tasks, Liu et al. proposed that the re-scaling DNN training tasks to fit gaps in schedules be formulated as a mixed-integer linear programming (MILP) problem, and demonstrated via simulation the potential benefits of the approach. Here, we introduce MalleTrain, a system that provides the first practical implementation of this approach and that furthermore generalizes it by allowing it to be used even for DNN training applications for which model information is unknown before runtime. Key to this latter innovation is the use of a lightweight online job profiling advisor (JPA) to collect critical scalability information for DNN jobs---information that it then employs to optimize resource allocations dynamically, in real time. We describe the MalleTrain architecture and present the results of a detailed experimental evaluation on a supercomputer GPU cluster and several representative DNN training workloads, including neural architecture search and hyperparameter optimization. Our results not only confirm the practical feasibility of leveraging idle supercomputer nodes for DNN training but improve significantly on prior results, improving training throughput by up to 22.3% without requiring users to provide job scalability information.
- Open Access
- 7 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 87-100
RAPIDS: Reconciling Availability, Accuracy, and Performance in Managing Geo-Distributed Scientific Data
In modern science, big data plays an increasingly important role. Many scientific applications, such as running simulations on supercomputers or conducting experiments on advanced instruments, produce huge amount of data at unprecedented speed. Analyzing and understanding such big data is the key for scientists to make scientific breakthroughs. However, data might become unavailable for scientists to access when outages or maintenance of the storage system occur, which severely hinders scientific discovery. To improve the data availability, data duplication and erasure coding (EC) are often used. But as the scientific data gets larger, using these two methods can cause considerable storage and network overhead. In this paper, we propose RAPIDS, a hybrid approach that combines the multigrid-based error-bounded lossy compression with erasure coding, to significantly reduce the storage and network overhead required for maintaining high data availability. Our experiments show that RAPIDS reduces the storage overhead by up to 7.5x and network overhead by up to 3x to achieve the same level of availability compared to the regular EC method. We improve RAPIDS by building two models to optimize the fault tolerance configurations and data gathering strategy. We demonstrate that RAPIDS significantly improves performance when running on many CPU cores in parallel or on GPUs.
- 7 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 87-100
Open AccessRAPIDS: Reconciling Availability, Accuracy, and Performance in Managing Geo-Distributed Scientific Data
In modern science, big data plays an increasingly important role. Many scientific applications, such as running simulations on supercomputers or conducting experiments on advanced instruments, produce huge amount of data at unprecedented speed. Analyzing and understanding such big data is the key for scientists to make scientific breakthroughs. However, data might become unavailable for scientists to access when outages or maintenance of the storage system occur, which severely hinders scientific discovery. To improve the data availability, data duplication and erasure coding (EC) are often used. But as the scientific data gets larger, using these two methods can cause considerable storage and network overhead. In this paper, we propose RAPIDS, a hybrid approach that combines the multigrid-based error-bounded lossy compression with erasure coding, to significantly reduce the storage and network overhead required for maintaining high data availability. Our experiments show that RAPIDS reduces the storage overhead by up to 7.5x and network overhead by up to 3x to achieve the same level of availability compared to the regular EC method. We improve RAPIDS by building two models to optimize the fault tolerance configurations and data gathering strategy. We demonstrate that RAPIDS significantly improves performance when running on many CPU cores in parallel or on GPUs.
- 31 March 2023
- Published by ASME International in Journal of Nondestructive Evaluation, Diagnostics and Prognostics of Engineering Systems
- Vol. 6 (4) , 1-13
Full Waveform Inversion-Based Ultrasound Computed Tomography Acceleration Using Two-Dimensional Convolutional Neural Networks
Ultrasound computed tomography (USCT) shows great promise in nondestructive evaluation and medical imaging due to its ability to quickly scan and collect data from a region of interest. However, existing approaches are a tradeoff between the accuracy of the prediction and the speed at which the data can be analyzed, and processing the collected data into a meaningful image requires both time and computational resources. We propose to develop convolutional neural networks (CNNs) to accelerate and enhance the inversion results to reveal underlying structures or abnormalities that may be located within the region of interest. For training, the ultrasonic signals were first processed using the full waveform inversion (FWI) technique for only a single iteration; the resulting image and the corresponding true model were used as the input and output, respectively. The proposed machine learning approach is based on implementing two-dimensional CNNs to find an approximate solution to the inverse problem of a partial differential equation-based model reconstruction. To alleviate the time-consuming and computationally intensive data generation process, a high-performance computing-based framework has been developed to generate the training data in parallel. At the inference stage, the acquired signals will be first processed by FWI for a single iteration; then the resulting image will be processed by a pre-trained CNN to instantaneously generate the final output image. The results showed that once trained, the CNNs can quickly generate the predicted wave speed distributions with significantly enhanced speed and accuracy.
- 31 March 2023
- Published by ASME International in Journal of Nondestructive Evaluation, Diagnostics and Prognostics of Engineering Systems
- Vol. 6 (4) , 1-13
Full Waveform Inversion-Based Ultrasound Computed Tomography Acceleration Using Two-Dimensional Convolutional Neural Networks
Ultrasound computed tomography (USCT) shows great promise in nondestructive evaluation and medical imaging due to its ability to quickly scan and collect data from a region of interest. However, existing approaches are a tradeoff between the accuracy of the prediction and the speed at which the data can be analyzed, and processing the collected data into a meaningful image requires both time and computational resources. We propose to develop convolutional neural networks (CNNs) to accelerate and enhance the inversion results to reveal underlying structures or abnormalities that may be located within the region of interest. For training, the ultrasonic signals were first processed using the full waveform inversion (FWI) technique for only a single iteration; the resulting image and the corresponding true model were used as the input and output, respectively. The proposed machine learning approach is based on implementing two-dimensional CNNs to find an approximate solution to the inverse problem of a partial differential equation-based model reconstruction. To alleviate the time-consuming and computationally intensive data generation process, a high-performance computing-based framework has been developed to generate the training data in parallel. At the inference stage, the acquired signals will be first processed by FWI for a single iteration; then the resulting image will be processed by a pre-trained CNN to instantaneously generate the final output image. The results showed that once trained, the CNNs can quickly generate the predicted wave speed distributions with significantly enhanced speed and accuracy.
- Open Access
- 25 October 2022
- Published by International Union of Crystallography (IUCr) in Journal of Synchrotron Radiation
- Vol. 29 (6) , 1495-1503
ROI-Finder: machine learning to guide region-of-interest scanning for X-ray fluorescence microscopy
The microscopy research at the Bionanoprobe (currently at beamline 9-ID and later 2-ID after APS-U) of Argonne National Laboratory focuses on applying synchrotron X-ray fluorescence (XRF) techniques to obtain trace elemental mappings of cryogenic biological samples to gain insights about their role in critical biological activities. The elemental mappings and the morphological aspects of the biological samples, in this instance, the bacterium Escherichia coli (E. Coli), also serve as label-free biological fingerprints to identify E. coli cells that have been treated differently. The key limitations of achieving good identification performance are the extraction of cells from raw XRF measurements via binary conversion, definition of features, noise floor and proportion of cells treated differently in the measurement. Automating cell extraction from raw XRF measurements across different types of chemical treatment and the implementation of machine-learning models to distinguish cells from the background and their differing treatments are described. Principal components are calculated from domain knowledge specific features and clustered to distinguish healthy and poisoned cells from the background without manual annotation. The cells are ranked via fuzzy clustering to recommend regions of interest for automated experimentation. The effects of dwell time and the amount of data required on the usability of the software are also discussed.
- 25 October 2022
- Published by International Union of Crystallography (IUCr) in Journal of Synchrotron Radiation
- Vol. 29 (6) , 1495-1503
Open AccessROI-Finder: machine learning to guide region-of-interest scanning for X-ray fluorescence microscopy
The microscopy research at the Bionanoprobe (currently at beamline 9-ID and later 2-ID after APS-U) of Argonne National Laboratory focuses on applying synchrotron X-ray fluorescence (XRF) techniques to obtain trace elemental mappings of cryogenic biological samples to gain insights about their role in critical biological activities. The elemental mappings and the morphological aspects of the biological samples, in this instance, the bacterium Escherichia coli (E. Coli), also serve as label-free biological fingerprints to identify E. coli cells that have been treated differently. The key limitations of achieving good identification performance are the extraction of cells from raw XRF measurements via binary conversion, definition of features, noise floor and proportion of cells treated differently in the measurement. Automating cell extraction from raw XRF measurements across different types of chemical treatment and the implementation of machine-learning models to distinguish cells from the background and their differing treatments are described. Principal components are calculated from domain knowledge specific features and clustered to distinguish healthy and poisoned cells from the background without manual annotation. The cells are ranked via fuzzy clustering to recommend regions of interest for automated experimentation. The effects of dwell time and the amount of data required on the usability of the software are also discussed.
- 3 October 2022
- Published by SPIE-Intl Soc Optical Eng
- Vol. 12227, 152-158
Assessment of unsupervised clustering algorithms for label-free x-ray fluorescence microscopy data
We previously demonstrated a machine learning based regions-of-interest (ROI) finding tool for X-ray fluorescence microscopy, called XRF-ROI-Finder at the 9-ID beamline in Argonne National Laboratory.1 Bacterial cell treatment type prediction and recommendation for steering experiments were performed via the application of fuzzy k-means clustering algorithm. ROI-Finder takes the fluorescence microscopy images, performs segmentation and detects individual E.coli cells, extracts features for principal component analysis, and ultimately performs label-free clustering for cell treatment type prediction and recommendation for similar cells to perform automatic steering experimentation. In this paper, we assess two additional clustering method, namely hierarchical agglomerative clustering (HAC) and density based spatial clustering of applications with noise (DBSCAN) algorithm. The ROI-Finder software is hosted at https://github.com/aisteer/ROI-Finder.
- 3 October 2022
- Published by SPIE-Intl Soc Optical Eng
- Vol. 12227, 152-158
Assessment of unsupervised clustering algorithms for label-free x-ray fluorescence microscopy data
We previously demonstrated a machine learning based regions-of-interest (ROI) finding tool for X-ray fluorescence microscopy, called XRF-ROI-Finder at the 9-ID beamline in Argonne National Laboratory.1 Bacterial cell treatment type prediction and recommendation for steering experiments were performed via the application of fuzzy k-means clustering algorithm. ROI-Finder takes the fluorescence microscopy images, performs segmentation and detects individual E.coli cells, extracts features for principal component analysis, and ultimately performs label-free clustering for cell treatment type prediction and recommendation for similar cells to perform automatic steering experimentation. In this paper, we assess two additional clustering method, namely hierarchical agglomerative clustering (HAC) and density based spatial clustering of applications with noise (DBSCAN) algorithm. The ROI-Finder software is hosted at https://github.com/aisteer/ROI-Finder.
- Open Access
fairDMS: Rapid Model Training by Data and Model Reuse
Extracting actionable information rapidly from data produced by instruments such as the Linac Coherent Light Source (LCLS-II) and Advanced Photon Source Upgrade (APS-U) is becoming ever more challenging due to high (up to TB/s) data rates. Conventional physics-based information retrieval methods are hard-pressed to detect interesting events fast enough to enable timely focusing on a rare event or correction of an error. Machine learning~(ML) methods that learn cheap surrogate classifiers present a promising alternative, but can fail catastrophically when changes in instrument or sample result in degradation in ML performance. To overcome such difficulties, we present a new data storage and ML model training architecture designed to organize large volumes of data and models so that when model degradation is detected, prior models and/or data can be queried rapidly and a more suitable model retrieved and fine-tuned for new conditions. We show that our approach can achieve up to 100x data labelling speedup compared to the current state-of-the-art, 200x improvement in training speed, and 92x speedup in-terms of end-to-end model updating time.
Open AccessfairDMS: Rapid Model Training by Data and Model Reuse
Extracting actionable information rapidly from data produced by instruments such as the Linac Coherent Light Source (LCLS-II) and Advanced Photon Source Upgrade (APS-U) is becoming ever more challenging due to high (up to TB/s) data rates. Conventional physics-based information retrieval methods are hard-pressed to detect interesting events fast enough to enable timely focusing on a rare event or correction of an error. Machine learning~(ML) methods that learn cheap surrogate classifiers present a promising alternative, but can fail catastrophically when changes in instrument or sample result in degradation in ML performance. To overcome such difficulties, we present a new data storage and ML model training architecture designed to organize large volumes of data and models so that when model degradation is detected, prior models and/or data can be queried rapidly and a more suitable model retrieved and fine-tuned for new conditions. We show that our approach can achieve up to 100x data labelling speedup compared to the current state-of-the-art, 200x improvement in training speed, and 92x speedup in-terms of end-to-end model updating time.
- Open Access
- 29 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Real-time porosity mapping and visualization for synchrotron tomography
Applications of X-ray computed tomography (CT) for porosity characterization of engineering materials often involve an extended data analysis workflow that includes CT reconstruction of raw projection data, binarization, labeling and mesh extraction. It is often desirable to map the porosity in larger samples but the computational challenge of reducing gigabytes of raw data to porosity information poses a critical bottleneck. In this work, we describe algorithms and implementation of an end-to-end porosity mapping code that processes raw projection data from a synchrotron CT instrument into a porosity map and visualization in the form of triangular face mesh. Towards this objective, we report the development of a novel subset reconstruction scheme for X-ray CT using filtered backprojection and a convolutional neural network that allows us to reconstruct arbitrarily-shaped subsets of a tomography object. We build upon this scheme to implement the complete code for porosity mapping. The code first detects possible voids by performing a coarse reconstruction on down-sampled projections and then improves the shape of those voids with higher detail offered by reconstructing selected subsets from the original raw data. We report measurements of the time taken by this code to perform complete processing from raw data to a triangular face mesh for several visualization scenarios on a single highperformance workstation equipped with GPU. We show that we can now visualize local porosity within a 8 gigavoxel CT volume (12 gigabytes raw data) within 1 to 2 minutes and a 64 gigavoxel CT volume (100 gigabytes of raw data) within 3 to 7 minutes.
- 29 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessReal-time porosity mapping and visualization for synchrotron tomography
Applications of X-ray computed tomography (CT) for porosity characterization of engineering materials often involve an extended data analysis workflow that includes CT reconstruction of raw projection data, binarization, labeling and mesh extraction. It is often desirable to map the porosity in larger samples but the computational challenge of reducing gigabytes of raw data to porosity information poses a critical bottleneck. In this work, we describe algorithms and implementation of an end-to-end porosity mapping code that processes raw projection data from a synchrotron CT instrument into a porosity map and visualization in the form of triangular face mesh. Towards this objective, we report the development of a novel subset reconstruction scheme for X-ray CT using filtered backprojection and a convolutional neural network that allows us to reconstruct arbitrarily-shaped subsets of a tomography object. We build upon this scheme to implement the complete code for porosity mapping. The code first detects possible voids by performing a coarse reconstruction on down-sampled projections and then improves the shape of those voids with higher detail offered by reconstructing selected subsets from the original raw data. We report measurements of the time taken by this code to perform complete processing from raw data to a triangular face mesh for several visualization scenarios on a single highperformance workstation equipped with GPU. We show that we can now visualize local porosity within a 8 gigavoxel CT volume (12 gigabytes raw data) within 1 to 2 minutes and a 64 gigavoxel CT volume (100 gigabytes of raw data) within 3 to 7 minutes.