Browse Prior Art Database

Apply Slow Close Process to Minimize Physical Open/Close Activities for Database Datasets

IP.com Disclosure Number: IPCOM000100276D
Original Publication Date: 1990-Mar-01
Included in the Prior Art Database: 2005-Mar-15
Document File: 3 page(s) / 163K

Publishing Venue

IBM

Related People

Haderle, DJ: AUTHOR [+3]

Abstract

This article describes a technique which can be used by a database system to efficiently manage its datasets' open/close activities. Before referencing and modifying data within a dataset, the dataset must be physically opened. Since data in a single dataset can be concurrently accessed by multiple transactions within a database system, the database system must provide functions to physically or logically (i.e., keep track the use count) open or close datasets.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 38% of the total text.

Apply Slow Close Process to Minimize Physical Open/Close Activities for Database Datasets

       This article describes a technique which can be used by a
database system to efficiently manage its datasets' open/close
activities.  Before referencing and modifying data within a dataset,
the dataset must be physically opened.  Since data in a single
dataset can be concurrently accessed by multiple transactions within
a database system, the database system must provide functions to
physically or logically (i.e., keep track the use count) open or
close datasets.

      Normally, a dataset will be physically closed when there are no
current users of the dataset.  For performance considerations, it is
desirable to have datasets stay open, once opened.  However, there is
always a physical limitation on the number of datasets that can be
concurrently opened within a database system.  Therefore, a facility
is normally provided by a database system to allow users to make
their own decision on which datasets need to stay open once opened.

      Frequently accessed datasets, once opened, can be kept open by
associating a CLOSE(NO) type attribute with those datasets.  A
CLOSE(NO) type dataset will be physically opened on the first
reference to the dataset.  Once it is opened, it will not be
physically closed until the database system is terminated or it is
explicitly closed by the user through a database command.

      For each dataset that is defined with a CLOSE(YES) type
attribute, the database system will maintain a global dataset
reference count.  The reference count will be incremented/decremented
whenever the dataset is logically opened/closed by a database
transaction.  Normally, a database transaction will perform only one
logical open against a referenced dataset within a COMMIT scope.  The
open dataset will be logically closed when the transaction is
committed.  The physical open will only occur on the first logical
open request.  A CLOSE(YES) type dataset will be physically closed
when the last logical close is requested (i.e., dataset reference
count goes to zero).

      Due to the physical limit on the number of concurrent open
datasets, it is very difficult for database administrators to
determine how many frequently accessed datasets need to be defined
with a CLOSE(NO) type attribute. If the number is set too high, then
it will have a higher chance of getting an open failure when the
system is heavily loaded.  On the other hand, if the number is set
too low, then database transactions will suffer from performance
drawback by constantly performing physical open/close of those
CLOSE(YES) type datasets.  Therefore, database administrators will
have a hard time correctly and efficiently managing the distribution
of CLOSE(YES or NO) type attributes among all their database
datasets.

      To alleviate the above problem, this invention introduces a
SLOW CLOSE process which efficiently manages the open/...