Converting from a database randomized with DFSHDC40

You can convert from a database that is randomized with DFSHDC40 into an HDAM database randomized with the Sequential Subset Randomizer.

Procedure

Converting an existing DFSHDC40-HDAM database into an HDAM database randomized with the Sequential Subset Randomizer requires the following steps:

  1. Generate a temporary Sequential Subset Randomizer.

    For this temporary Sequential Subset Randomizer, the database administrator does not need to define accurately how much space should be allocated to each subset. However, all other specifications need to be accurate. Note that at this time the DBDGEN should not be changed.

    This temporary Sequential Subset Randomizer is required so that the SS-STATS routine produces statistics for each subset. The SS-STATS routine finds the description of all subsets (including the values of the subset IDs) in the tables of the temporary Sequential Subset Randomizer to make statistics.

  2. Unload the database using FABHURG1 or FABHFSU with the SSSTATS control statement specified (see Figure 1 for an example).

    The database unload will produce the following:

    • An unloaded database which will be used as input for the reload.
    • Statistics that describe how much space should be allocated to each subset. These statistics should be used for a second, definitive SSRGEN which will be used during the reload process.

    It is recommended that you activate the Database Tuning Statistics during this database unload. (See Obtaining statistics for database tuning.) The Database Tuning Statistics will show whether the database is well organized and whether the current randomizing is efficient. If the current randomizing is inefficient, the database administrator can take corrective actions before reloading the database. For example, if the root addressable area is too crowded, the database administrator can increase the size of the root addressable area. The Database Tuning Statistics can also be compared with the Database Tuning Statistics created at a later time, when the database is randomized with the Sequential Subset Randomizer. This can be used to see whether the Sequential Subset randomizing is as efficient as DFSHDC40 randomizing.

  3. Generate the final version of the Sequential Subset Randomizer.

    For this second Sequential Subset Randomizer, the database administrator should provide accurate specifications about the relative amount of space to be allocated to each subset. The relative amount of space to be allocated to each subset can be found in the SS-STATS statistics in the previous database unload.

    If the database administrator knows in advance that some subsets will eventually incur an overproportional increase in data volume, the data administrator can use this knowledge in order to increase the relative amount of space allocated to that subset. This may avoid a later adjustment of Sequential Subset Randomizer tables and an associated database reorganization.

  4. Change the DBD source statements in order to specify the name of the generated Sequential Subset Randomizer as the randomizer.

    This is a good time to review the current randomizing parameters and make adjustments with existing Database Tuning Statistics or other statistics if necessary. You can review the number of CIs/blocks in the root addressable area, the number of RAPs per CI/block, the CI/block size, the bytes limit, the scan parameter, and the free space specifications.

    Then the DBDGEN must be performed and the ACBGEN should be performed if necessary.

  5. Create sorted input for the database reload.

    For small databases with less than 10,000 roots, this step is not necessary.

    For large databases, the input to the reload should be provided in the correct RAP sequence of the new database in order to complete the reload in a reasonable amount of time. For conversions from DFSHDC40 to Sequential Subset Randomizer, the correct sequence for the reload can be created without lengthy sort by executing the FABIUNLS utility (see Figure 1 for an example). The FABIUNLS utility splits the unloaded data set into multiple data sets so that one subset becomes one data set.

    Therefore, the input to the reload will consist of a concatenation of the multiple output data sets of FABIUNLS.

  6. Reload the database with the new or modified DBD using the Sequential Subset Randomizer.
  7. Obtain Database Tuning Statistics to check whether the quality of the randomizing is appropriate.

    This can be done by unloading a database, activating the Database Tuning Statistics for this unload, and setting the output of the unload to dummy.