wmo et-adrs: 23-25 april 2008 1 wmo et-adrs hierarchical data format (hdf) manuel fuentes (ecmwf)...
TRANSCRIPT
WMOET-ADRS: 23-25 April 2008 1
WMO ET-ADRSHierarchical Data Format (HDF)
Manuel Fuentes
(ECMWF)
Erdem Erdi
(Turkish State Meteorological Service)
WMOET-ADRS: 23-25 April 2008 2
Outline
Brief introduction to HDF
SWOT Analysis
Practical examples
WMOET-ADRS: 23-25 April 2008 3
Hierarchical Data Format: HDF
HDF is a file format
HDF files are self-described
HDF technologies at present include two data management formats (HDF4 and HDF5) and libraries, a modular data browser/editor, associated tools and utilities, and a conversion library
Both HDF4 and HDF5 were designed to be a general scientific format, adaptable to virtually any scientific or engineering application, and also have been used successfully in non-technical areas
HDF5 is particularly good at dealing with data where complexity and scalability are important
WMOET-ADRS: 23-25 April 2008 4
Features provided by HDF5 technology
Unlimited size, extensibility, and portability
General data model
Unlimited variety of datatypes
Flexible, efficient I/O
Flexible data storage
Data transformation and complex subsetting
WMOET-ADRS: 23-25 April 2008 5
HDF5: Unlimited size, extensibility, and portability
HDF5 does not limit the size of files or the size or number of objects in a file.
The HDF5 format and library are extensible and designed to evolve gracefully to satisfy new demands.
HDF5 functionality and data is portable across virtually all computing platforms and is distributed with C, C++, Java, and Fortran90 programming interfaces.
WMOET-ADRS: 23-25 April 2008 6
HDF5: General data model
The HDF5 data model supports complex data relationships and dependencies through its grouping and linking mechanisms.
HDF5 accommodates many common types of metadata and arbitrary user-defined metadata.
WMOET-ADRS: 23-25 April 2008 7
HDF5: Unlimited variety of datatypes
HDF5 supports a rich set of pre-defined datatypes as well as the creation of an unlimited variety of complex user-defined datatypes.
Datatype definitions can be shared among objects in an HDF file, providing a powerful and efficient mechanism for describing data.
Datatype definitions include information such as byte order (endian), size, and floating point representation, to fully describe how the data is stored, insuring portability to other platforms.
WMOET-ADRS: 23-25 April 2008 8
Data model and datatypes
WMOET-ADRS: 23-25 April 2008 9
HDF5: Flexible, efficient I/O
HDF5, through its virtual file layer, offers extremely flexible storage and data transfer capabilities. Standard (Posix), Parallel, and Network I/O file drivers are provided with HDF5.
Application developers can write additional file drivers to implement customized data storage or transport capabilities.
The parallel I/O driver for HDF5 reduces access times on parallel systems by reading/writing multiple data streams simultaneously.
WMOET-ADRS: 23-25 April 2008 10
HDF5: Flexible data storage
HDF5 employs various compression, extensibility, and chunking strategies to improve access, management, and storage efficiency.
HDF5 provides for external storage of raw data, allowing raw data to be shared among HDF5 files and/or applications, and often saving disk space.
WMOET-ADRS: 23-25 April 2008 11
HDF5: Data transformation and complex subsetting
HDF5 enables datatype and spatial transformation during I/O operations.
HDF5 data I/O functions can operate on selected subsets of the data, reducing transferred data volume and improving access speed.
WMOET-ADRS: 23-25 April 2008 12
Governance: The HDF Group
The mission of The HDF Project is
to develop, promote, deploy and support open and free technologies that facilitate scientific data exchange, access, analysis, archiving and discovery
to ensure long-term availability and support for HDF technologies, and by extension, long-term accessibility of data stored using HDF technologies
The HDF group currently includes 15 full time staff members and 3 to 5 students. The group’s annual budget is $2.1 million, which is mostly provided by the government sector
WMOET-ADRS: 23-25 April 2008 13
Copyright
http://hdf.ncsa.uiuc.edu/HDF5/doc/Copyright.html
HDF5 (Hierarchical Data Format 5) Software Library and Utilities with Copyright 2006-2008 by The HDF Group (THG).
NCSA HDF5 (Hierarchical Data Format 5) Software Library and Utilities with Copyright 1998-2006 by the Board of Trustees of the University of Illinois.
All rights reserved.
WMOET-ADRS: 23-25 April 2008 14
Copyright (cont.)
Redistribution and use in source and binary forms, with or without modification, are permitted for any purpose (including commercial purposes) provided that the following conditions are met:
Redistributions of source code must retain the above copyright notice, this list of conditions, and the following disclaimer
Redistributions in binary form must reproduce the above copyright notice (which is on the previous slide) , this list of conditions, and the following disclaimer in the documentation and/or materials provided with the distribution
In addition, redistributions of modified forms of the source or binary code must carry prominent notices stating that the original code was changed and the date of the change
All publications or advertising materials mentioning features or use of this software are asked, but not required, to acknowledge that it was developed by The HDF Group and by the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign and credit the contributors
Neither the name of The HDF Group, the name of the University, nor the name of any Contributor may be used to endorse or promote products derived from this software without specific prior written permission from THG, the University, or the Contributor, respectively
DISCLAIMER: THIS SOFTWARE IS PROVIDED BY THE HDF GROUP (THG) AND THE CONTRIBUTORS "AS IS" WITH NO WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED. In no event shall THG or the Contributors be liable for any damages suffered by the users arising out of the use of this software, even if advised of the possibility of such damage
WMOET-ADRS: 23-25 April 2008 15
SWOT Analysis: Criteria
Ability to present information pertinent to WMO Programmes
Ability to encode textual information, such as warnings
Ability for usage in operational data exchanges
Ability for usage in transmission of information to users outside NMHSs
Ability for usage in storage systems by NMHSs, centres or other users
Compliance and status with regard to existing standards
Inter-operability, translation back and forward to other DRSs
Can it be used to envelope objects
Available and widespread support (skills and technology)
WMOET-ADRS: 23-25 April 2008 16
SWOT : Present information pertinent to WMO
Ability/suitability to present information pertinent to WMO Programmes and Member needs including weather, climate, water, atmospheric constituents, oceanography, aviation and other related environmental information
Any data of 2-D, 3-D meteorology, hydrology or similar science can be handled
Used by satellite applications in meteorology due to suitability for large and complex data.
There are not too many tools for non-satellite meteorological data
It’s not clear how to handle millions of bulletins:
Group bulletins into a big file
1 bulletin per file (minimum HDF5 file size: 2 Kbyte)
WMOET-ADRS: 23-25 April 2008 17
SWOT : Present information for pictorial display
Ability/suitability to present information for pictorial display
HDF can store and present graphical data with 2 or 3 dimensions, allows for raster and vectors. Tools can display information in graphical form
WMOET-ADRS: 23-25 April 2008 18
SWOT : Encode textual information
Ability/suitability to encode textual information, such as warnings
HDF can store textual information of any length
Suitable for storing metadata
WMOET-ADRS: 23-25 April 2008 19
SWOT : Encode Metadata
WMOET-ADRS: 23-25 April 2008 20
SWOT : Usage in operational data exchanges
Ability/suitability for usage in operational data exchanges (real time or otherwise) between NMHSs and centres. Including information regarding existing usage especially with regard to extent of use
EUMETSAT dissemination (EUMETCAST) supports HDF as delivery format
There is no naming convention for satellite data (only TERRA & AQUA share naming convention, because the same team developed the 2 satellites). Otherwise each satellite has different naming convention and order of elements in file
There are 2 attempts to standardize satellite data in HDF:
• KNMI-HDF5: Special library for encoding
• HDF-EOS: TERRA, AQUA & Petabytes more
WMOET-ADRS: 23-25 April 2008 21
SWOT : Usage in operational data exchanges
An important portion of the operational satellite based meteorological data and products are distributed to the meteorological community in the HDF format, in near-realtime or non-realtime (archive).
Some examples are
EUMETSAT SAF Products (NWC SAF, LAND SAF)
EUMETSAT EPS data
NASA EOS Data and products
WMOET-ADRS: 23-25 April 2008 22
SWOT : Transmission of information to users outside NMHS
Ability/suitability for usage in transmission of information to users outside NMHSs or centres. Including information regarding existing usage especially with regard to extent of use
HDF is widely used in scientific communities:
• Universities, Research labs
• Space agencies (like NASA and EUMETSAT)HDF is mainly used for satellite data.
Use of HDF in a variety of disciplines and users
• Encourages development of tools
• Makes it easy to use outside NMHSsSoftware publicly available with supported tools
WMOET-ADRS: 23-25 April 2008 23
SWOT : Usage in storage systems
Ability/suitability for usage in storage systems by NMHSs, centres or other users. Including information regarding existing usage especially with regard to extent of use
Parallel I/O
Machine independent
Compression
The HDF Group is committed to ensure the long-term accessibility of HDF-stored data
EUMETSAT does not store data in HDF, but convert from raw
NASA archives the Earth Observing System data in HDF
“Grouping” of data at archiving may impose restrictions on how data can be retrieved
WMOET-ADRS: 23-25 April 2008 24
SWOT : Standards
Compliance and status with regard to with existing standards. Are they open standards? Which body overseas them. Is there any proprietary nature to them. Are they flexible enough to accommodate our current and foreseen needs. How are they updated, is it a straight forward process
HDF format is very suitable for GIS (as it can handle both data and metadata in the same file). However, it is not widely used for GIS because of the lack of a convention (schema)
It is governed by The HDF Group
Compression: SZIP method is proprietary, ZLIB is open
HDF licence seems flexible
The library is updated regularly in a straight forward manner
WMOET-ADRS: 23-25 April 2008 25
SWOT : Interoperability
How suitable is the DRS to the WIS and to developing the appropriate metadata? Is existing documentation good? How much variance is there in current implementations? Are the existing flavours inter-operable?
HDF can meet the requirements regarding metadata required for the WIS
Documentation is good with lots of examples
There are 2 implementations:
Tools to convert from HDF4 to HDF5
No direct inter-operability between the 2 implementations
WMOET-ADRS: 23-25 April 2008 26
SWOT : Conversion to/from other DRS
What are the issues for translating back and forward to other DRSs?
Translation could be loss-less when using same encoding method (not standard)
Encoding: Offset and scale factor. Tools are not aware of encoding
Compression can be used instead of encoding in order to avoid larger files
Compression is transparent for users
Native data types are 1, 2, 4, 8 bytes
WMOET-ADRS: 23-25 April 2008 27
SWOT : Envelope objects
Can they be used to envelope objects or act as a pseudo-carrier for other data formats?
HDF can handle/envelope any kind of data format either binary or ASCII
HDF can handle BLOBs (stream of bytes)
WMOET-ADRS: 23-25 April 2008 28
SWOT: Support, skills and technology
Available and widespread support for the DRS (skills and technology)
The HDF Group’s commitment to:
• Support HDF
• Ensure long-term accessibility of the dataEstablished user community
WMOET-ADRS: 23-25 April 2008 29
SWOT Summary: Strengths
HDF5 can store and present 2-D or 3-D data (gridded fields), together with metadata
Rich set of predefined datatypes and data relationships
High performance features:
Parallel I/O
Unlimited dimensions
Compression
Unlimited size and amount of data
I/O functions can operate on subsets of data
Open data format and free software (libraries and tools)
Operational services (EUMETCAST) support HDF
WMOET-ADRS: 23-25 April 2008 30
SWOT Summary: Weaknesses
HDF is a file format, as opposed to a message/bulletin format (like GRIB or BUFR)
There is no convention:
Names to use
Order in which to store elements
May not handle well point observation data
There aren’t many tools for meteorological data
Comparison with NetCDF:
HDF5’s general data model makes writing data more difficult than NetCDF
HDF5 will be the storage format for NetCDF4
WMOET-ADRS: 23-25 April 2008 31
SWOT Summary: Opportunities
Using HDF5 may improve inter-operability with other disciplines
Using HDF5 may improve usability of meteorological data outside NMHSs
Software publicly available, with numerous (general) tools and programming languages
WMOET-ADRS: 23-25 April 2008 32
SWOT Summary: Threats
The HDF format is developed and maintained by a single group (The HDF Group). Any problem with funding could jeopardise the existence of the format or its support
Meteorology would be very small community compared to other users of HDF. Requirements of the Meteorological community may not be so important for the HDF community
WMOET-ADRS: 23-25 April 2008 33
Practical Examples
Data received at ECMWF, converted to BUFR, then used by Forecasting System:
HDF4
Microwave Brightness Temperature from Tropical Rainfall Measuring Mission (trmm)
Rainfall from Tropical Rainfall Measuring Mission (trmm)
HDF5
METOP GOME-2 total column ozone data
Aura OMI ozone data
Each data stream has its own conversion tool HDF to BUFR
WMOET-ADRS: 23-25 April 2008 34
Practical Examples
Eumetsat
HDF-EOS
HDF-KNMI
We haven’t found any examples of field observations (SYNOP, METAR, radiosondes..)
WMOET-ADRS: 23-25 April 2008 35
Practical Example: MODIS swath data from channel 3
WMOET-ADRS: 23-25 April 2008 36
Practical Example: 3-D data
Schwarzschild metric (spatial components only)
WMOET-ADRS: 23-25 April 2008 37
Practical Example: MSG total precipitable water
WMOET-ADRS: 23-25 April 2008 38
Practical Example: SAF cloudtype product
WMOET-ADRS: 23-25 April 2008 39
Conclusion
HDF5 is going to be the base for NetCDF4. It makes more sense to focus on NetCDF4 than HDF5
Support for subsetting
Parallel I/O
Unlimited dimensions
Compression
Remove current limitations on file size
HDF is a file format as opposed to GRIB/BUFR which are message (or bulletin) formats
HDF might not be suitable for operational exchange of meteorological data between NMHSs, but to present meteorological information to other users/disciplines