Monitoring group buffer pool delete name requests

You can determine how many times delete-name requests occur in a group buffer pool.

To determine how many times delete-name requests occur in a group buffer pool, use one of the following approaches:

• Issue the following DISPLAY GROUPBUFFERPOOL command with the MDETAIL option.
  For example, you might issue the following command:

  -DISPLAY GROUPBUFFERPOOL(GBP29) MDETAIL(*)

  The following example shows partial output of the DISPLAY GROUPBUFFERPOOL command, which includes delete-name information:

  DSNB750I - DISPLAY FOR GROUP BUFFER POOL GBP29 FOLLOWS
         .
         .
         .
  DSNB797I - OTHER INTERACTIONS
             REGISTER PAGE                               = 0
             UNREGISTER PAGE                             = 0
             DELETE NAME                                 = 0
             READ STORAGE STATISTICS                     = 0
             EXPLICIT CROSS INVALIDATIONS                = 0
             ASYNCHRONOUS GBP REQUESTS                   = 0
         .
         .
         .
  DSNB793I - DELETE NAME LIST                            = 0
             READ CASTOUT STATISTICS                     = 0
             DELETE NAME                                 = 0
             OTHER ASYNCHRONOUS GBP REQUESTS             = 0
  DSNB790I - DISPLAY FOR GROUP BUFFER POOL GBP29 IS COMPLETE
  DSN9022I - DSNB1CMD '-DISPLAY GBPOOL' NORMAL COMPLETION

• Use the IBM® OMEGAMON® for Db2 Performance Expert on z/OS® Statistics Detail report or IFCID 263 (Page set castout detail).

Example

Assume that you issue the following command:

DISPLAY GROUPBUFFERPOOL (GBP0) MDETAIL

The following example shows a partial output from the command:

⋮
DSNB796I -DB1A CASTOUTS
               PAGES CASTOUT                             = 217
               UNLOCK CASTOUT                            = 35
               READ CASTOUT CLASS                        = 47
               READ CASTOUT STATISTICS                   = 47
               READ DIRECTORY INFO                       = 290
⋮

The UNLOCK CASTOUT counter should always be significantly less than the PAGES CASTOUT counter. If, at the very least, it is not less than half, the castout write I/O is not performed efficiently. (The number of pages written per I/O is normally close to the number that is obtained by dividing PAGES CASTOUT by UNLOCK CASTOUT). This is probably because you have random update patterns on the Db2 data.