On the Design and Scheduling of an Index Processing System for Very Large Databases

One of the major operations in an inverted file or indexed retrieval system is the combining of two or more lists of pointers to items in the database, based on a Boolean expression, to form composite lists of pointers of interests. While this is a simple operation, it is not performed efficiently by a general purpose digital computer. In this thesis, we present the organization and operation of an index processing system which offers speed increases in performing this operation. The effects of the design parameters, such as merge network configuration, disk storage system architecture, query complexity and arrival rate, and postings list length and overlaps, on the performance of this system are examined, using both analytical techniques and simulations. Five algorithms for scheduling the merge trees are also defined and their effects on system performance studied. Results of these studies will allow the implementor of a very large text database to select the system configuration and scheduling algorithms which would produce the greatest performance under a specific users software environment.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2014

Author dc:creator

• Huang, Hui-Ming

• Computer Science

Language dc:language

eng

Identifier

(UMI)AAI8108547

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/66444