SQL/MP Reference Manual

HP NonStop SQL/MP Reference Manual—523352-013
P-6
Specifying Configuration for Parallel Index Loading
3 * Number of Records / Number of Partitions
SQL estimates the number of records in the base table by dividing the file size by
the record length. The estimate of the number of records is used by the sort
process to calculate the scratch file size. For more information on the scratch file
size, see the FastSort Manual.

Swap file—The swap file for a sort process defaults to the same volume as the
scratch volume if the scratch volume is local. If the scratch volume is not local, the
swap file defaults to $SYSTEM. The swap file for a record generator process is the
volume of the partition being read if the partition is local. If the partition is not local,
the swap volume is the default swap volume from the =_DEFAULTS DEFINE.
Specifying Configuration for Parallel Index Loading
The CONFIG option allows you to specify an EDIT file that contains a description of a
default configuration or an explicit configuration for both record generators and sort
processes.
The configuration file can contain two types of configuration statements: comments and
CREATEINDEX statements. Keywords in the configuration file can be in uppercase,
lowercase, or mixed-case letters.
The COMMENT statement includes comments in the file. SQL ignores all lines that
begin with the word COMMENT or the characters ==. The syntax is:
With the CREATEINDEX option you can specify:

Default priority for the record generators and the sort processes (PRI)

Default object files for the record generators and the sort processes (PROGRAM)

Default number of records (NUMRECS)

Default pool of processors in which to run the record generators and another pool
in which to run the sort processes (processor)

Default pool of volumes to use for the initial set of sort scratch files for the sort
processes (SCRATCH)

Default pool of volumes to use for overflow storage for the sort processes, if
needed (SCRATCHON)
Note. If no configuration information is present, all record generators that read remote
partitions swap to the same volume, and all sort processes that write to remote partitions
swap to $SYSTEM. Multiple processes swapping to the same volume might cause disk
space and contention problems on that volume so you might want to specify a
configuration for parallel index loading.
{ COMMENT comment-text }
{ == comment-text }