Related Experiment Videos
Feature-based Analysis of Large-scale Spatio-Temporal Sensor Data on Hybrid Architectures
Joel Saltz1, George Teodoro1, Tony Pan1
1Center for Comprehensive Informatics and Biomedical Informatics Department Emory University.
The International Journal of High Performance Computing Applications
|May 13, 2017
Summary
This study introduces middleware to analyze large sensor datasets efficiently. It leverages hybrid CPU-GPU clusters for data-intensive feature analysis in complex spatio-temporal data.
Area of Science:
- Data science
- High-performance computing
- Scientific computing
Background:
- Large sensor datasets are crucial for fields like climate modeling and biomedicine.
- Analyzing this data is computationally expensive and faces scalability challenges.
- Current methods struggle with the volume and complexity of spatio-temporal data.
Purpose of the Study:
- To develop middleware for efficient analysis of large sensor datasets.
- To enable feature-based analysis on large spatio-temporal data.
- To overcome computational barriers in big data analysis.
Main Methods:
- Utilizing large clusters of hybrid CPU-GPU nodes.
- Implementing middleware system support for data-intensive tasks.
- Focusing on feature-based analysis techniques.
Main Results:
- Demonstrated middleware system support for hybrid CPU-GPU clusters.
- Addressed data and compute-intensive requirements for feature analysis.
- Facilitated analysis of large spatio-temporal datasets.
Conclusions:
- Middleware systems can effectively harness hybrid CPU-GPU resources.
- This approach overcomes barriers in analyzing large sensor data.
- Enables advanced feature extraction in diverse scientific domains.
