Access HDF Data in the Cloud via OPeNDAP Web Service
Download as PPTX, PDF•0 likes•60 views
The document discusses accessing HDF data in the cloud using DMRPP, highlighting performance improvements with direct I/O for faster response times. It covers the successful mapping and access of NASA HDF4 and HDF-EOS2 products, and mentions the need for certain DMRPP files to be regenerated to utilize the direct I/O feature effectively. Additionally, it notes ongoing work to enhance geolocation data storage from HDF-EOS2/HDF4.
Access HDF Data in the Cloud via OPeNDAP Web Service
- 1. 2024 ESIP Summer Meeting Accessing HDF Data in the Cloud via OPeNDAP Web Service Kent Yang Software Engineer/NASA EED-3 contractor [email protected] GOVERNMENT RIGHTS NOTICE This work was authored by employees of The HDF Group under Contract No. 80GSFC21CA001 with the National Aeronautics and Space Administration. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, worldwide license to reproduce, prepare derivative works, distribute copies to the public, and perform publicly and display publicly, or allow others to do so, for United States Government purposes. All other rights are reserved by the copyright owner. ©2024 Raytheon Company. All rights reserved.
- 2. Topics Overview ● Accessing HDF* Data in the Cloud via dmrpp** ● Direct IO*** Performance Improvement ● Work in progress to access NASA HDF4 and HDF-EOS****2 files *Hierarchical Data Format ** Dataset Metadata Response Plus Plus *** Input Output **** Earth Observing System
- 3. Direct IO Performance Improvement Concept HDF5 File dmrpp File NetCDF NetCDF* File Decompress Compress Hyrax Core Pass through the data Pass through the data HDF5 File dmrpp File NetCDF NetCDF File Hyrax Core * Network Common Data Form General Approach Approach with Direct IO
- 4. Hyrax Server Response Time Speed-up With Direct IO Product Sample File File Size (MB) Response Time without Direct IO (Seconds) Response Time with Direct IO (Seconds) Speed-up in Response Time by using Direct IO GHRSST* 9 2.8 0.2 14 X TROPOMI** 292 26.6 1.8 15 X SSMI*** 1.4 0.5 0.3 1.7 X *: Group for High Resolution Sea Surface Temperature **: TROPOspheric Monitoring Instrument ***: Special Sensor Microwave Imager
- 5. Big Files With Direct IO Product Sample File File Size (GB) Response Time with Direct IO (Second) Server Response Message Without using Direct IO Daymet 3.5 45 The maximum response time limit(165 seconds) is exceeded. MODIS* Derived 4 65 Insufficient memory CH4 Level 4 11 Direct IO feature is not used because it doesn’t contain any compressed variable. The maximum response time limit(165 seconds) is exceeded. * Moderate Resolution Imaging Spectroradiometer
- 6. Facts for the Direct IO Feature ● Hyrax will use direct IO automatically for those cases when end users request to obtain the whole array of the selected variable(s) and those variable(s) are compressed. ● This process is entirely transparent to the end users. ● Direct IO doesn’t work for some old dmrpp files if they don’t contain the key information needed for using the Direct IO feature. These dmrpp files need to be regenerated to take advantage of the Direct IO feature.
- 7. Direct IO Performance Improvement Summary ● Can greatly reduce server computation time ● Can greatly reduce server memory usage ● Use HDF5 direct chunk IO API*s ● The feature is in the current Hyrax release * Application Programming Interface
- 8. Accessing HDF4 and HDF-EOS2 via dmrpp ● Map HDF4 to DMR* ● Access HDF4 via dmrpp ○ Not only handle data stored in chunking and contiguous layouts ○ Also need to handle data stored in linked blocks ○ Handle HDF-EOS2/HDF4 geolocation data ■ Not stored as HDF4 variables ■ Need to calculate them based on the metadata information ■ Save the data in a proper way * Dataset Metadata Response
- 9. Current Status ● We can successfully map and access the sample NASA HDF4 and HDF- EOS2 products via dmrpp. ● We are still working on a better way to store HDF-EOS2/HDF4 geolocation data.
- 10. ● Panoply screenshots of variable Topography ○ The identical plots show the dmrpp module can successfully access this HDF4 file AIRS* Local HDF4 file via dmrpp’s netCDF-4 file Local HDF4 file netCDF-4 file via dmrpp * Atmospheric Infrared Sounder
- 11. Thank you!
- 12. This work was supported by NASA/GSFC under Raytheon Company contract number 80GSFC21CA001