Database performance evaluation in an indexed file environment

Abstract

The use of database systems for managerial decision making often incorporates information-retrieval capabilities with numeric report generation. Of great concern to the user of such a system is the response time associated with issuing a query to the database. This study presents a procedure for estimating response time for one of the most frequently encountered physical storage mechanisms, the indexed file. The model provides a fairly high degree of accuracy, but is simple enough so that the cost of applying the model is not exorbitant. The model incorporates the knowledge that the distribution of access key occurrences is known to follow Zipf's law. It first estimates the access time required to complete the query, which includes the time needed for all input and output transactions, and CPU time used in performing the search. The effects of multiple users on an individual's response time are then assessed using a simple regression estimation technique. The two-step procedure allows for the separation of access time from multiuser influences.