Browse Prior Art Database

Performance Prediction for Secondary Index Files

IP.com Disclosure Number: IPCOM000077250D
Original Publication Date: 1972-Jun-01
Included in the Prior Art Database: 2005-Feb-25
Document File: 2 page(s) / 56K

Publishing Venue

IBM

Related People

Ling, H: AUTHOR [+2]

Abstract

In formatted file information system design, the designer often implements secondary index files on those fields when they think those fields will be queried. The use of this technique, however, does not always give an improvement. Sometimes it gives degradation. The use of the predictor as described here will give a recommendation whether or not to implement the secondary index files.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 65% of the total text.

Page 1 of 2

Performance Prediction for Secondary Index Files

In formatted file information system design, the designer often implements secondary index files on those fields when they think those fields will be queried. The use of this technique, however, does not always give an improvement. Sometimes it gives degradation. The use of the predictor as described here will give a recommendation whether or not to implement the secondary index files.

The figure shows the implementation procedure step-by-step. The sample data base logical structure and the representative transaction mix are fed into stage A of the predictor. For each field F(i), all the possible costs of indexing it are generated in block A. Here F(ij) is the probability that field Fi can be retrieved through the indexing of the field F(j), and C(ij) is the space costs for having field F(j) indexed. These data are sent to the matrix stores blocks B and C. In block C, C(iR) represents the cost of retrieval of field F(i), C(iI) represents the cost of insertion of field F(i), and C(iu) represents the cost of update of field F(i). The storage can be cores, registers or any storage device. The products of these data are calculated and stored in block D according to the matrix multiplication shown as follows: X(i) represents the extra cost that can be saved if field F(i) is not indexed, Y(i) represents the extra cost spent in insertion if field F(i) is indexed and Z(i) represents the extra cost spent in update maintenan...