Journaling Procedures › Influencing Journaling Performance › Improving Warmstart Performance

Improving Warmstart Performance

Reducing Warmstart Time

You can reduce the time it takes to warmstart a central version following an abnormal termination by specifying a non-zero value for a JOURNAL FRAGMENT INTERVAL in the system generation SYSTEM statement or in a DCMT VARY JOURNAL command.

How the Journal Fragment Works

The journal fragment interval designates an interval for writing dummy segment (DSEG) records to the journal file. DC/UCF uses the DSEG records in the event of a system crash to determine the appropriate starting place for warmstart processing, as shown in the following steps:

1. The new journal file is activated. It begins with header records. These records contain:
   • Information on currently open transactions
   • The relative block number (RBN) of the DSEG record. The RBN signifies which DSEG record is used to start forward processing in the event of a warmstart.

   

2. If the journal fragment interval is 500, the DC/UCF system will do the following before it writes the 509th journal block:
   • Creates and writes the DSEG record
   • Updates the DSEG RBN in the journal header

   

3. In the event of a system crash, the warmstart forward processing starts at the DSEG record at RBN 509 instead of at the JSEG record. This saves the time it would have taken for processing to read the first 500 journal blocks.

   

Considerations

If your journal files are large (in terms of the number of pages), a journal fragment interval can significantly reduce the amount of time it takes to warmstart a DC/UCF system. The warmstart logic goes to the most recently accessed journal fragment and starts its recovery processing from that point. However, because there is overhead required to write dummy segment headers, your journal fragment interval should be at least 100. Choose an interval that is between 100 and half the number of blocks in your journal file.