To collect tuning information, it is necessary to collect the statistical information (statistics log) output by HiRDB. The statistics log is output to a statistics log file. The HiRDB administrator must determine an appropriate size for this file.
There are two statistics log files, named pdstj1 and pdstj2, under the $PDDIR/spool directory. HiRDB creates these statistics log files automatically. The HiRDB administrator must specify the size of the statistics log file in the pd_stj_file_size operand. The following points should be noted about determining this file size:
Table 20-2 Servers subject to collection of statistics log information
Type of statistics log | HiRDB/Single Server | HiRDB/Parallel Server | ||||
---|---|---|---|---|---|---|
SDS | MGR | FES | DS | BES | ||
System activity statistical information | sys | Y | Y | Y | Y | Y |
UAP statistical information | uap | Y | ![]() | Y | ![]() | ![]() |
SQL statistical information | sql | Y | ![]() | Y | ![]() | ![]() |
Global buffer pool statistical information | buf | Y | ![]() | ![]() | Y | Y |
Statistical information on HiRDB files for database manipulation | fil | Y | ![]() | ![]() | Y | Y |
Deferred write processing statistical information | dfw | Y | ![]() | ![]() | Y | Y |
Index statistical information* | idx | Y | ![]() | ![]() | Y | Y |
SQL static optimization information | sop | Y | ![]() | Y | ![]() | ![]() |
SQL dynamic optimization information | dop | Y | ![]() | Y | ![]() | ![]() |
SQL object statistical information | pcd | Y | ![]() | Y | Y | Y |
SQL statement statistical information | sqh | Y | ![]() | Y | ![]() | ![]() |
Statistics on SQL object transfer | obj | ![]() | ![]() | ![]() | Y | Y |
Statistical information on the operation of foreign servers | fsv | ![]() | ![]() | ![]() | ![]() | Y |
Statistical information on the usage status of foreign servers | hba | ![]() | ![]() | ![]() | ![]() | Y |
Y: Statistics log information that is collected
: Statistics log information that is not collected
* Information on index page splits is not collected as part of the statistics log. For information on how to collect this information, see 20.2 Collecting tuning information from the system log.
When a statistics log file becomes full during output of statistics log information, HiRDB switches outputs of the information to the other statistics log file. This is called statistics log file swapping. When statistics log file swapping occurs, new information is written over any existing statistics log information in the swapped-in file. For this reason, as soon as a statistics log file is swapped out, the HiRDB administrator should unload it and create an unload statistics log file. Figure 20-1 shows the procedure for swapping the statistics log files.
Figure 20-1 Swapping of statistics log files