Browse Prior Art Database

Counter for Full Table Scan Control in Data Redistribution

IP.com Disclosure Number: IPCOM000109648D
Original Publication Date: 1992-Sep-01
Included in the Prior Art Database: 2005-Mar-24
Document File: 2 page(s) / 68K

Publishing Venue

IBM

Related People

Chen, WT: AUTHOR [+2]

Abstract

Disclosed is a method of using a counter to determine if a full table scan operation can stop. In a loosely coupled multinode database system, data records are partitioned into buckets and all the data records belonging to a bucket should always be placed into a single node. If data redistribution requires to move a bucket from its original residence to another node, every data record of this particular bucket will be moved to its new residing node. However, most relational database systems do not support physical bucket in their data storage organizations, the data redistribution program has to run a full table scan to fetch the records that belong to the buckets for redistribution. Because a full table scan is usually costly, it has a significant influence on the performance of the entire data redistribution operation.

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

Counter for Full Table Scan Control in Data Redistribution

       Disclosed is a method of using a counter to determine if
a full table scan operation can stop. In a loosely coupled multinode
database system, data records are partitioned into buckets and all
the data records belonging to a bucket should always be placed into a
single node. If data redistribution requires to move a bucket from
its original residence to another node, every data record of this
particular bucket will be moved to its new residing node.  However,
most relational database systems do not support physical bucket in
their data storage organizations, the data redistribution program has
to run a full table scan to fetch the records that belong to the
buckets for redistribution. Because a full table scan is usually
costly, it has a significant influence on the performance of the
entire data redistribution operation.

      In data redistribution, the buckets to be moved are
predetermined.  As the number of data records of each bucket is
known, the total number of data records to be retrieved from a
particular table can be derived.  For example, let's assume Bucket 10
and 23 of table Branches are currently residing in Node 1 and they
are the only buckets of table Branches in Node 1 to be moved to Node
998.  It is known that there are 123 records in Bucket 10 and 345
records in Bucket 23.  Thus, the data redistribution should retrieve
totally 468 records from table Branches in Node 1 for sending to Node
998.  When the data redistribution program runs a full table sca...