The significant difference between batch and On-Line applications
is the amount of automation needed to process a transaction. Batch
processes must rely on mechanical techniques to make decisions concerning
the fate of a transaction record whereas On-Line systems depend on
human interaction.
Performance criterion for On-Line systems revolves around response
time. The limiting factor relating to response time is the degree of
I/O performed per name inquiry. NameSearch? makes it possible to cluster
records on name keys yielding optimal utilization of I/0 resources.
In many cases it is affordable to use large search ranges and eliminate
unlikely candidates using the comparison routine. This method allows
the greatest degree of name variations to be processed and displays
only the most likely candidates. The extra CPU expense incurred by
comparing the records in a larger set is nominal, since only a single
entry per transaction is being processed.
Similar to On-Line systems, batch utilities, which create reports
or merge files, need to be miserly with I/O resources. Typically, batch
applications evaluate large input files against the database. The sequence
of records being processed can greatly effect the amount of page swapping
being managed by the database system. By including NameSearch? ranges
and sorting the transaction file based on the start value, the number
of page swaps will be minimized and the optimum I/O utilization yield
will be maximized..
Unlike On-Line systems, the cost of processing large ranges of candidates
is very CPU intensive. Each record in the transaction file must be
compared with all the records returned for that set. As the sets get
larger, the number of comparisons per transaction record increases.
The result of increasing the number of comparisons per record is prohibitive
when calculating the total number of comparisons performed on the entire
transaction file. Batch processes must limit the size of name search
sets in order to avoid becoming CPU bound.
In both batch and On-Line systems the sizes of ranges, use of comparison
routines, and database organizations are determined by the size of
the database being searched, performance objectives, and expectations
relating to accuracy and the expense of implementation. As the size
of your database grows these factors become harder to balance. It is
extremely important to use a realistic test environment to tune your
name search application. An approach that worked well for a small data
sample will perform differently on your production system.
Technical Product Information
