Browse Prior Art Database

Management of Modified Pages Using VM Data Spaces by SQL/DS

IP.com Disclosure Number: IPCOM000121451D
Original Publication Date: 1991-Sep-01
Included in the Prior Art Database: 2005-Apr-03
Document File: 2 page(s) / 74K

Publishing Venue

IBM

Related People

Fecteau, G: AUTHOR [+3]

Abstract

Disclosed is a method that will allow SQL/DS* to reduce the checkpoint overhead when using VM/ESA* Data Spaces, as well as to provide overlap of database writes with other database activities.

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

Management of Modified Pages Using VM Data Spaces by SQL/DS

      Disclosed is a method that will allow SQL/DS* to reduce
the checkpoint overhead when using VM/ESA* Data Spaces, as well as to
provide overlap of database writes with other database activities.

      When SQL/DS modifies data in a data space, the operating system
is allowed (but not required) to rewrite the updated copy to the
database disk. The 'MAPMDISK SAVE' interface is provided to request
that the disk copy be made current. SQL/DS ensures that all modified
data is current on disk at checkpoint time. In order to avoid issuing
all the 'MAPMDISK SAVE' requests at checkpoint time, a count of
modified pages and a bit map is kept in memory. A threshold is set as
the 'SAVE' request interval. When this threshold is  reached, a
'SAVE' request is issued, reducing the number of 'SAVEs' that will be
required at checkpoint time to a manageable number.

      SQL/DS uses the asynchronous  'SAVE' interface. A parameter is
used to identify each 'SAVE' request by a unique value.  Control is
returned to SQL/DS before performing the I/O required by the 'SAVE',
thus processing can overlap I/O. An interrupt is presented once the
I/O for the 'SAVE' has completed, returning the unique value passed
during the 'SAVE' request. Data can be modified between the 'SAVE'
request and 'SAVE' completion. In that case, a subsequent 'SAVE' will
be issued.

      Checkpoint completion ensures that all data has been written to
disk. Therefore, SQL/DS waits for all outstanding 'SAVEs' to complete
before marking the checkpoint as completed.

      To avoid issuing 'SAVE' request for an entire Data Space,
SQL/DS keeps a bit map made of 1 bit for every 32 pages (one block)
of Data Spaces (called the DSMOD bit). Every time a page is modified,
the co...