Browse Prior Art Database

High Performance Large Object Database

IP.com Disclosure Number: IPCOM000123825D
Original Publication Date: 1999-May-01
Included in the Prior Art Database: 2005-Apr-05
Document File: 2 page(s) / 69K

Publishing Venue

IBM

Related People

Ciarlante, JJ: AUTHOR

Abstract

A program is disclosed that will allow for easy data manipulation (store, retrieve, replace, delete, and backup).

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 58% of the total text.

High Performance Large Object Database

   A program is disclosed that will allow for easy data
manipulation (store, retrieve, replace, delete, and backup).

   Porting an MVS mainframe application to an AIX RISC/6000
platform presented performance and processor utilization problems.
The application reads and stores image data from a high speed scanner
at a rate of 80 images/second.  No commercial database we tested was
able to sustain this rate without using 100% of the RISC/6 000
processor.

   To solve the problem the AIX file system and its utilities
are used in a unique way.

   Begin by strategically placing 100 numbered directories
(/00 to /99) on multiple physical disks to be accessed sequentially.
Each data object will be stored as a standard AIX file.

   Files stored in the database will be numbered sequentially
and stored in the directory that corresponds to the last 2 digits of
its sequence number.  The name of the file consists of the first part
(high order) of its sequence number written in reverse.  The rest of
the file name can be anything meaningful to the application.

   This database design takes advantage of both hardware
and software on a small system to achieve its performance
objectives.  To take advantage of the hardware the directories are
spread evenly across the available disks, going from disk to disk to
alleviate disk arm contention.  The sequence numbers in the file
names are reversed to take advantage of the software.  Each time a...