Browse Prior Art Database

Page-Copying Method for Data Redistribution in Parallel Database System

IP.com Disclosure Number: IPCOM000106180D
Original Publication Date: 1993-Oct-01
Included in the Prior Art Database: 2005-Mar-20
Document File: 2 page(s) / 102K

Publishing Venue

IBM

Related People

Ballard, DJ: AUTHOR [+7]

Abstract

Disclosed is a method that can improve the receiving node's performance in storing received data into tables during data redistribution in a parallel database system. It provides a page copying method, which assumes the data records in the communication buffer arriving at the receiving node are put in the same format, except the header and tail area, as in the physical file created for database table storage. By handling multiple data records a time, the page copying method makes the data insertion with faster performance.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Page-Copying Method for Data Redistribution in Parallel Database System

      Disclosed is a method that can improve the receiving node's
performance in storing received data into tables during data
redistribution in a parallel database system.  It provides a page
copying method, which assumes the data records in the communication
buffer arriving at the receiving node are put in the same format,
except the header and tail area, as in the physical file created for
database table storage.  By handling multiple data records a time,
the page copying method makes the data insertion with faster
performance.

      Large volume data redistribution usually occurs when new nodes
are added into the parallel database system.  Data records to be
redistributed to the newly added nodes are concurrently collected in
each existing node and are sent to the new nodes for storage into the
table to which they belong.  The amount of data to be handled in the
new nodes (receiving nodes) is usually much larger than the amount
handled in each of the existing nodes (sending nodes), with the
assumption that the number of sending nodes is larger than the number
of receiving nodes.  Thus, the performance improvement on the
receiving nodes is worth pursuing, even if a slight overhead may be
added on the sending nodes.

      In data transmissions, data is put in a communications buffer.
The communications buffer is divided into three areas: (1) the header
that stores communication control information, (2) the data area that
stores the data records, and (3) the tail.  The table to which these
data records belong can be specified in previous messages or in the
communications buffer header.  Usually, when the receiving node
receives the communications buffer, it will directly or indirectly
read in the data records stored in the buffer and use the data record
insertion facility provided by the database system to insert each
record into the specified table.  The physical storage of a database
table can be considered as a data file containing pages and each page
containing data records.  The data record insertion facility locates
an existing page in the data file or allocates a new page to the data
file for inserting a data record.  The record insertion facility is
provided by the Data Management Service (DMS) in a database system.
DMS requires each page in a specific format that divides the page
into: (1) a header for storing information about the page itself, (2)
a da...