Abstract:
|
Advances in computing has outpaced progress in bandwidth storage and the ability to quickly read data from disk. Scientists are able to perform large-scale, high-resolution simulations, but are unable to examine all the data at once and queries can be prohibitively slow. Bitmap indexing is a technique to improve the speed of queries and basic data analysis. Complex logical operations can be performed quickly when data is indexed with bitmaps, but for floating-point attributes, the index will be lossy because the bit vector values must be binned. In addition, binning may be used to minimize the space requirements of the index. Many binning strategies have been proposed; however, a cohesive assessment method to compare the indexes is lacking. This talk will focus on the development of assessment methods to assist researchers in choosing an appropriate binning strategy for representing their raw data with a binned bitmap index.
|