Database Reference
In-Depth Information
usually includes the size of the dataset; name of the dataset variable; co-
ordinate system used in the dataset (spherical, Cartesian, etc.); the schema
(structured, unstructured, etc.); centering (cell centered, vertex centered, edge
centered, etc.); and number of coordinates that must exist per centering ele-
ment (each vertex, each face, etc.).
The various
VariableDescriptor
groups are then organized under one TOC
(table of contents) group that retains common global information about the
file's variables (the names of all variables, bitmap indexes, metadata infor-
mation). For the raw datasets, a unique
TimeStep
group is created for each
timestep in the discrete time range. Under each
TimeStep
group exists one
HDF5 dataset that contains the raw data for a given variable at that timestep.
The bitmap dataset corresponding to the variable is also stored under the same
TimeStep
group.
This division between data and metadata is essential for the primary reason
that variable metadata for a given dataset is relevant and accurate across
all timesteps for that dataset variable (there is no need to store redundant
metadata). Figure 9.9 illustrates the HDF5 FastQuery architectural layout.
9.4.3 Status and Sample Applications
H5Part research and development is an active, international collaborative ef-
fort involving researchers from high-performance computing and accelerator
modeling. It has evolved from focusing on I/O for particle-based datasets
HDF5-FastQuery Architecture
HDF5 ROOT Group
/HDF5_UC
TOC
Variable data
Variable descriptors
Descriptor
for
Variable 0
Descriptor
for
Variable 1
Descriptor
for
Variable X
…
…
Time Step 0
Time Step 1
Time Step Y
X variable descriptors
for X datasets
D0
D0
D0
Symbol Key
HDF5 Group:
Contains user retrievable information
about sub-groups and datasets
HDF5 Dataset:
Contains the actual data array of a given
variable X at a time step Y
D1
D1
D1
…
DX
…
DX
…
DX
Figure 9.9
Architectural layout of HDF5 FastQuery.
