The Advancing Art of Indexing - High-Performance Web Databases: Design, Development, and Deployment - page 647

Exhibit 46-2. Performance comparison of a query with and without
inverted indexes.
Benchmarks were performed against various data file structures, includ-
ing relational and flat file. In this case, a query performed on a 1-million
record data base needed to find all customers in Texas who ordered in the
previous month. This query required a cross-table join, based on a free-for-
mat address field in the CUSTOMER file and a date range in the ORDER file.
The results are shown in Exhibit 2 and demonstrate that inverted indexing
can guarantee consistent performance enhancements for an organization's
data access requirements.
SUMMARY
In the rush to serve the constantly expanding demands of knowledge
workers, businesses have created a complex environment for IS to devel-
op and maintain enterprisewide data access services. Parallel process-
ing, multidimensional servers, and partitioning are all brute-force
methods proposed to address data access performance and query flexi-
bility. Alternatives that minimize I/O and maximize memory processing
may deliver the best performance for the price. Inverted file indexing may
be a relatively simpler and cost-effective solution for many businesses.
These multiple keyword indexes allow users to perform ad hoc queries
with minimal impact to online systems. Users are also able to construct
complex queries quickly. In addition, by providing feedback to the user
on the size of their request before data is retrieved, client/server network
traffic is minimized.
Inverted indexing leverages investments in existing hardware and soft-
ware, allowing for the integration of new technologies while protecting
much of the application developed. Inverted indexes provide the broadest
range of flexibility for providing true data access across the enterprise.
Sometimes, simple is better.