Database Reference
In-Depth Information
hold true, and at worst, we are virtually guaranteed the viewer will not be
able to gain any meaningful information from the visual information overload.
If we scale up our dataset from gigabyte (10
9
) to terabyte (10
12
), we then
can expect on the order of between 199 million and 9.5 billion triangles repre-
senting a depth complexity ranging between about 80 and 4,700, respectively.
Regardless of which estimate of the number of triangles we use, we end up
drawing the same conclusion: depth complexity, and correspondingly scene
complexity and human cognitive workload, all grow at a rate that is a func-
tion of the size of the source data. Even if we are able to somehow display
all those triangles, we would be placing an incredibly dicult burden on the
user. They will be facing the impossible task of trying to visually locate smaller
needles in a larger haystack.
The multifaceted approach we are adopting takes square aim at the fun-
damental objective: help scientific researchers more quickly and eciently do
science. In one view, one primary tactical approach that seems promising is
to help focus user attention on easily consumable images from the large data
collection. We do not have enough space in this chapter to cover all aspects
in this regard. Instead, we provide a few details about a couple of especially
interesting challenge areas.
9.5.1 Implementing Query-Driven Visualization
In principle, the QDV idea is conceptually quite simple: restrict visualization
and analysis processing only to data of interest. In practice, implementing
this capability can be quite a challenge. Our approach is to take advantage of
the state-of-the-art index/query technology mentioned earlier in this chapter,
FastBit, and use it as the basis for data subsetting in query-driven visualiza-
tion applications. In principle, any type of data-subsetting technology can be
used to provide the same functionality. In practice, FastBit has proven to be
very ecient and effective at problems of scale.
Some of our early work in this space focused on comparing the performance
of FastBit as the basis for index/query in QDV applications with some of
the industry standard algorithms for isosurface computation.
34
In computing
isosurfaces, one must first find all the grid cells that contain the surface of
interest, then for each such cell, compute the isosurface that intersects the
cell. Our approach, which leverages FastBit, shows a performance gain of
between 25% to 300% over the best isocontouring algorithms created by the
visualization community. Of greater significance is the fact that our approach
extends to
n
-dimensional queries (i.e., queries of
n
different variables), whereas
the indexing structures created for use in isocontouring are applicable only to
univariate queries. This approach was demonstrated on datasets created by
a combustion simulation (see Figure 9.13, which shows three different query
conditions performed on a single dataset; see Stockinger et al.
34
for more
details).
