Performance data collection

With any software release, the support for new functions can cause an incremental change (increase or decrease) in performance for some workloads.

IBM makes a significant effort to minimize negative effects, and in most cases a change in performance is not observed. However, in a skip-release upgrade, there is a higher potential that the compound effect of incremental performance differences will be observed.

To understand the performance characteristics of an upgrade, collect performance data both before and after the upgrade is complete. Following are two recommended performance collections:

• Collect general performance data for one or more days by using Collection Services. Gathering performance data for benchmark contains information on ways to start Collection Services.
• Collect specific profile data for approximately an hour by running the following program in a batch job during a period of high activity on the server. To call this program, enter the command CALL SYSPROF PARM(PEXDTAVxRy) where x and y indicate the i5/OS™ version and release numbers.
  Note: You cannot store PEX data from two different releases into the same library.

  SYSPROF:  PGM PARM(&LIB)                                 
  DCL VAR(&LIB) TYPE(*CHAR) LEN(10)  /* Library to store the data */
  
  /* Create a PEX definition to collect profile data.       */
  ADDPEXDFN DFN(SYSPROF) TYPE(*TRACE) JOB((*ALL)) +     
               TASK(*ALL) MAXSTG(1000000) INTERVAL(5) +  
               TRCTYPE(*PRFDTA)                          
                                                         
  /* If the definition already exists, assume it is ok.     */
  MONMSG MSGID(CPFAF10)                              
                   
  STRPEX SSNID(SYSPROFILE) DFN(SYSPROF)                  
  DLYJOB 3600                                            
  ENDPEX SSNID(SYSPROFILE) DTALIB(&LIB)                  
  ENDPGM                                                 

Save this data from both collections for future reference.