Project Title:
A Query System for Heterogeneous, Distributed Database Management Using a Massively
92-1-07.04-8100 NAS09-18853
A Query System for Heterogeneous, Distributed
Database Management Using a Massively
Parallel Hyper-Index
Data Parallel Systems, Inc.
4617 East Morningside Drive
Bloomington, IN 47408
Latha S. Colby (812-334-8100)
Abstract:
Performance problems associated with accessing data that is
distributed among various sites, computers, and relational
database management systems are not amenable to purely software
solutions. The relative performance of commercial databases on a
variety of queries and hardware systems, the relative size and
location of tables, and network bandwidth and loading are issues
that must be addressed in order to balance loads and produce
reasonable response times. These factors also increase the
complexity of associated software. Even if all of these parameters
can be taken into account, current database technologies do not
provide the performance needed for such very large databases as
those maintained by NASA sites. The company has developed the
Hyper-Index Parallel Database Accelerator to accelerate query
processing on single relational databases by up to three orders-
of-magnitude, using the massive degree of parallelism of the
MasPar MP-1 supercomputer and other proprietary storage structures
and algorithms. The hyper-index is potentially a very powerful
network management tool for distributed databases because it can,
for many queries, use a limited amount of centralized data and
indexes to produce fast results while reducing network traffic.
Potential Commercial Application:
Potential Commercial Applications: A network-based, distributed
database management system, one that delivers the query
performance and the reduced network load offered by the hyper-
index approach, is a product for which government and business
will find many applications.
***