(Object) Relational Databases for HEP Data

Paper: 365
Session: C (talk)
Speaker: Athanas, Michael, University of Florida, Ithaca
Keywords: data bases, ODBMS's, object-oriented methods

Title: (Object) Relational Databases for HEP Data

Authors: Michael Athanas

Affiliation: University of Florida at Gainesville

Collaboration: Nile/CLEO

Commercial databases have been around in one form or another for quite some
time. Their evolution has been driven primarily by the needs of the business
community. The information explosion caused by the internet has motivated
new features from databases manufacturers such as the ability to manage
Gbytes to Tbytes of data and the ability to provide advanced abstract data
representation. The relevant question addressed by this paper is whether the
commercial database solution is a viable one for HEP.

We explore representation, population, and performance of commercial
databases in conjunction with real HEP data from the CLEO experiment. We
have developed a unified object representation of a database based upon STL.
We describe two implementations of this object for the Illustra Object
Relational database and the Informix Online Data Server relational database.
Due to the multi-dimensional nature of HEP data we have found data
representation in an object relational database convenient. Throughput
benchmark measurements have been performed where throughput is defined as
the quantity of "interesting" data per second. We discuss how the
conventional database client/server architecture adapts to a parallel
distributed environment as well as how function shipping may be used in the
open architecture provided by Illustra. We describe our overall experience
with the Illustra and Informix products.