2. What is a Datablade?
Image
HDF-EOS
Spatial
DataBlade API
Text
Web
Extending the server by
adding DataBlades: new Types
and new functions
Extensible
OR Engine
Scaleable
Data
Manager
INFORMIX-Universal Server
http://heineken.gsfc.nasa.gov/hdfeos/
3. What does the HDF-EOS Datablade
provide?
• Inventory management of large volume data and
metadata
• Flexible access to the data as a file or BinaryLarge-Object inside the database.
• Cross-granule searching and subsetting
• Re-use other datablades such as spatial search
tools from Informix
• Implements Grid subsetting operations in the
Database
http://heineken.gsfc.nasa.gov/hdfeos/
4. Plans for the Second Year
•
•
•
•
Implement Swath, Point, HDF datatypes
General subsetting tool: stored queries
Metadata search and query tools
Implement MISR Quality-Assessment database
http://heineken.gsfc.nasa.gov/hdfeos/
5. HDF-EOS Datablade Software Architecture
Informix DB
SQL
SQL
InterInterface
face
and
API
HDF-EOS
C library
(Unchanged)
Valuable
Data
Exposed
in Searchable
Tables
File
I/O
Emulator
http://heineken.gsfc.nasa.gov/hdfeos/
Granules
in
File
System
Granules
in
Informix
Features of the HDFEOS Datablade Software Architecture
The HDF-EOS library, although non-reentrant, still runs unchanged in the database, with no code modifications.
The File-I/O-Emulator (FIOE) catches HDF-EOS I/O calls, and is optimized to handle subsetting I/O patterns. Potential to optimize for other patterns as well.
The FIOE allows us to manage the data inside the database, or allow visibility to the data outside. User is not forced into one world or the other. Advantage of putting it inside is that you get extra safety features such as transaction management and recovery. Legacy science software can continue to run on the data if left outside.
Can expose valuable indexable data inside the database as necessary. (Vdata example)
So how does this architecture perform?….