Table 3-1 Collection level determined by the value of the pd_prf_level operand and the information that is collected

Specification guidelines

• We recommend that you specify 0000001f (default level) so that when you investigate performance-related errors, you can quickly identify the cause of the problem. If a decrease in online performance is acceptable, such as during environment configuration, consider specifying 0000007f (detail level).
• Normally there is no need to specify 000001ff (maintenance level). Specify this level only if you are requested to do so by the support service.
• Do not specify 00000000 (suppression level) because, in the event of an error, this setting makes it difficult to determine its cause.

Notes

• If the number of performance trace collection points is increased and the collection level is changed to acquire more detailed performance traces, online performance might decrease.
• Do not specify a value other than 00000007, 0000001f, 0000007f, 000001ff, or 00000000. If any other value is specified, it becomes more difficult to verify performance and to investigation errors.

Operand default value

If this operand is omitted, the value specified for the same operand in the system common definition takes effect. If the same operand is also omitted from the system common definition, 00000007 is assumed.

Relationship to other operands

This operand is related to the following operands:

• pd_prf_file_count
• pd_prf_file_size

Remarks

By using the pdprflevel command, you can change the collection level even while HiRDB is running.

Specification guidelines

• Normally, this operand does not need to be specified.
• Consider specifying this operand if you want to retain the information in the performance trace information file for a long time. Estimate the number of performance trace information file generations and the file size from the following total capacity of the performance trace information file:
  Total capacity of the performance trace files = pd_prf_file_count operand value pd_prf_file_size operand value
  Example
  This example specifies the total capacity of the performance trace information files required for retaining one hour worth of performance trace information.
  

  pd_prf_level operand value	Total capacity of performance trace information files
  	Number of SQL statements executed per second is about 200	Number of SQL statements executed per second is about 500
  00000007	750 MB (default value#)	750 MB (default value#)
  0000001f	750 MB (default value#)	2 GB
  0000007f	2 GB	5 GB

  #: This is default value of the pd_prf_file_count operand default value of the pd_prf_file_size operand.
  
• To approximate the total capacity of the performance trace information files required per unit, use the following guidelines:
  Total capacity of the performance trace information files = MAX(768,000, a b c d) (kilobytes)
  a: Number of front-end and back-end servers per unit
  For a HiRDB/Single Server or if there is only a system manager or a dictionary server, the value is 1.
  b: Number of SQL statements executed per second on each server
  You can obtain the average value per editing time by dividing TOTAL, the total SQL statement execution count in the SQL-related statistical information, by the editing time EDIT TIME.
  c: Average amount of performance trace collected per SQL statement (kilobytes)
  Assume one of the following values, based on the value of the pd_prf_level operand:
  00000007: 0.5 kilobytes
  0000001f: 0.5 kilobytes
  0000007f: 1 kilobyte
  Note that the above value varies according to conditions, such as the type of SQL statement, amount of data, system configuration, whether facilities are applied, and tuning status.
  d: Length of time for which the performance trace information is to be retained (seconds)
  
• The length of time for which the performance trace information can be retained per file might change drastically depending on numerous factors, such as the type of transaction and HiRDB configuration. To obtain accurate values, measure the time for which performance trace information is retained per file when the most typical transactions and utilities are executed. You can determine the time value by using the timestamp or the pdprfed command's editing results. Calculate the required number of file generations based on the measured time.

Notes

• This operand value affects the amount of disk space required under the HiRDB directory. The following amount of disk space is required to store the performance trace information files:
  pd_prf_file_count operand value pd_prf_file_size operand value
  
• This operand affects the capacity of the error information files that are collected by using commands such as pdinfoget.
• If you change the value of this operand to a smaller value, the performance trace information files corresponding to the eliminated portion of the value will no longer be used, but they are not deleted automatically. If you need to delete any unneeded performance trace information file, delete them after HiRDB has been terminated.

Operand default value

If this operand is omitted, the value specified for the same operand in the system common definition takes effect. If the same operand is also omitted from the system common definition, 15 is assumed.

Relationship to other operands

This operand is related to the pd_prf_file_size operand.

Specification guidelines

Normally, this operand does not need to be specified. Consider specifying this operand if you want to retain the performance trace information files for a longer than normal time.

Notes

• This operand value affects the amount of disk space required under the HiRDB directory. The following amount of disk space is required to store the performance trace information files:
  pd_prf_file_count operand value pd_prf_file_size operand value
  
• This operand affects the capacity of the error information files that are collected by using commands such as pdinfoget.
• If you increase this operand value, the time required by the pdprfed command to edit and output performance traces increases.
• If this operand value is too large, sending the files to the support service for purposes such as an investigation of performance issues might be difficult, even if the files are compressed. Specify the appropriate file size taking into account file transmission.
• The performance trace information is output to the performance trace information file every 10 seconds. If many traces are output at one time, the performance trace information file might become smaller in size by a maximum of 10 megabytes.

Operand default value

If this operand is omitted, the value specified for the same operand in the system common definition takes effect. If the same operand is also omitted from the system common definition, 51200 is assumed.

Relationship to other operands

This operand is related to the pd_prf_file_count operand.