Title:   Efficient Native XML Storage
Speaker: Vagelis Hristidis (Florida International University)

Abstract:

XML has emerged as one of the popular data-representation
formats for information storage and exchange. XML data
today range from representing small files to encapsulating
gigabytes of information. For persistence as well as cost-
efficiency, large XML databases must be stored on disk drives,
the most cost-effective medium today for mass storage of
data. Current approaches that map XML data to relational
databases or simply use flat files incur a mismatch between
the structure of XML data and the underlying storage de-
vice (disk drive). We study a new method to store XML
data on disk drives that matches the characteristics of XML
data with those of disk drives. In particular, we present
algorithms that, given an XML document and a disk drive,
decide how to store the document on the drive in a way that
will later allow efficient navigation on the XML data, which
leads to efficient execution of XML queries. We evaluate
our proposed method using the DiskSim disk simulator and
benchmark XPath queries. The experimental results indi-
cate significant savings in query execution time compared to
the baseline sequential layout of the XML data.