The Advancing Art of Indexing - High-Performance Web Databases: Design, Development, and Deployment - page 638

• Pull the analytical/historical data into a data warehouse, so that the
queries do not affect online production performance.
DSS inquiries and reports require access to large amounts of data, even
when only a small subset of records is of interest. Consequently, placing
limits on queries is generally unsatisfactory for all involved. The time
required to accomplish the DSS request depends on disk throughput rather
than CPU speed. While CPU speeds continue to increase and disk capaci-
ties see great gains, the transfer speeds (disk I/O) have only made incre-
mental gains. So, users continue to be frustrated about the limits that are
set, and IS is again put in the position of “policing” access to data.
Offloading data into a data warehouse only shifts the problem from one
data structure to another. The assumption made with moving the data into
a query-focused data base is that the data warehouse, unlike a production
system, does not require the same response times. Knowledge workers
must wait hours, and sometimes days, for information they need.
The solution is to reduce the amount of disk I/O required to get the job
done. In data base applications, this means minimizing the amount of I/O
needed to select and retrieve a desired subset of records for display or re-
porting. With this in mind, indexing becomes one of the most important as-
pects of any decision support system.
INDEXING — THE FUTURE IS NOW
Decision support applications require users to query, analyze, and re-
port data. As data structures increase in size (millions and billions of
records of data), the ability to meet the need to query, analyze, and report
on data becomes burdensome — even for a super-powered computer.
Whether a DSS application is in the conceptual stage or already developed,
the issue of making data easily and immediately accessible to users will al-
ways be a challenge. Indexing provides a way to realize optimal benefits
with minimal investment in new technologies or equipment.
Sophisticated indexing is the most effective way to reduce the disk I/O
required to retrieve a subset of data. With advanced indexing techniques,
record selections by any criteria are accomplished using few disk reads. As
a result, complex selections from large data bases execute in seconds.
Not the New Kid on the Block, But Effective
Data file structures offer several ways to access data. Foremost among
them, sequential searches, or table scans, match data to the user's criteria.
This technique requires access to every record and, consequently, large
disk I/O. If available, an index can expedite this process by decreasing the
number of reads.