If you are using coupling facility structures for your log streams, you must specify in the CFRM policy the size of each structure you define. To simplify the task of calculating the size, IBM® provides the CF Structure Sizer (CFSizer), a tool that estimates the storage size for each structure by asking questions based on your existing or planned configuration. The CF Structure Sizer (CFSizer) uses the selected structure input to calculate the SIZE and INITSIZE values for the CFRM policy.

The IBM CF Structure Sizer (CFSizer) can be accessed at http://www.ibm.com/systems/support/z/cfsizer/.

Before using the CF Structure Sizer (CFSizer), you should go through the following steps for each coupling facility structure you wish to define. The information that you collect will be useful both for the CF Structure Sizer (CFSizer) and for defining your LOGR policy. The procedure will help you determine the minimum amount of coupling facility space you should allocate for a coupling facility structure, based on the amount of space needed for all the log streams mapping to the structure. See Add information about log streams and coupling facility structures to the LOGR policy.

1. Gather information about the coupling facility structure.
   For each structure, gather the following information common for all the log streams that map to the structure.

   AVGBUFSIZE

   Average size of log blocks written by applications to log streams associated with this coupling facility structure. This value corresponds to the AVGBUFSIZE parameter specified for each structure in the LOGR policy. For this procedure, pick the value that represents the average size of log blocks generated by all the log streams writing to this coupling facility structure. For specifying the AVGBUFSIZE parameter in the LOGR policy, see Specifying the average log block size.

   For a logrec log stream, IBM recommends an AVGBUFSIZE of 4068.

   For an OPERLOG log stream, IBM recommends an AVGBUFSIZE of 512.

   For an RRS log stream, see z/OS MVS Programming: Resource Recovery.

   Installation or vendor supplied applications should base AVGBUFSIZE on the average size of the log blocks they want to write to the log stream.

   MAXBUFSIZE

   Maximum size of a log block corresponding to the MAXBUFSIZE parameter in the LOGR policy.

   For a logrec or OPERLOG log stream, IBM recommends a MAXBUFSIZE of 4096. For an RRS log stream, see z/OS MVS Programming: Resource Recovery.

   Installation or vendor supplied applications should base MAXBUFSIZE on the maximum size of the log blocks they want to write to the log stream.

   Note that once you have defined the MAXBUFSIZE for a coupling facility structure, you cannot update the value. To change the MAXBUFSIZE, you must delete the log streams associated with the structure, delete the structure, and then re-define the structure with the new MAXGBUFSIZE value.

   LOGSNUM

   Maximum number of log streams in the structure corresponding to the LOGSNUM parameter in the LOGR policy. See The LOGSNUM parameter for more information.

2. Gather information about each log stream

   For each log stream that maps to a coupling facility structure, gather the following information.

   Most of these values are projected or desired values for your log stream, and a few correspond to parameters you will specify in the LOGR policy.

   LOWOFFLOAD

   LOWOFFLOAD specifies the point, in percent value of space consumed, where system logger will stop offloading coupling facility log data to the DASD log data sets for this log stream. It corresponds to the LOWOFFLOAD parameter in the LOGR policy. For OPERLOG, logrec log stream, or the CICS® log manager, IBM recommends that you use the LOWOFFLOAD default value of zero. See Administrative data utility for more the LOWOFFLOAD parameter.

   HIGHOFFLOAD

   HIGHOFFLOAD specifies the point, in percent value of space consumed, where system logger will begin offloading coupling facility log data to the DASD log data sets for this log stream. It corresponds to the HIGHOFFLOAD parameter in the LOGR policy.

   For OPERLOG, logrec log stream, or the CICS log manager, IBM recommends that you use the HIGHOFFLOAD default value of 80. See Administrative data utility for more the HIGHOFFLOAD parameter.

   ResidencyTime

   Desired residency time for log data in the coupling facility structure, in seconds. Residency time is the amount of time you want a log block to stay in the coupling facility between the time it is written to the coupling facility and being off-loaded to DASD.

   For a logrec or OPERLOG log stream, IBM recommends a residency of 10 seconds. For other system logger applications, get the desired residency time from the documentation for the application.

   WritePerSec

   Projected write requests per second against the log stream. This value is the total IXGWRITE requests per second issued by all system logger applications or systems connected to a log stream.

   For an OPERLOG log stream, you can calculate projected writes per second using the current SYSLOGs. For each system in the sysplex, do the following calculation:

   1. Choose a spot in SYSLOG and note the timestamp at that spot. You might want to choose a peak or high usage time of day.
   2. Page down 2000 lines by issuing, for example, DOWN 2000 to SDSF or TSO/E browse.
   3. Note the time stamp in this spot.
   4. Calculate the number of seconds between the two time stamps. This is the number of seconds it took to write 2000 lines of SYSLOG. (This calculation is based on most messages being one line).
   5. Divide 2000 by the number of seconds it took to write 2000 lines to get the lines per second.
   6. Add the results for each system's SYSLOG to get the total writes per second for the OPERLOG log stream.

   For a logrec log stream, calculate the projected writes per second using the current logrec data sets. For each system that will write to the logrec log stream, do the following calculation:

   1. Request or obtain an EREP report for a particular time span. It must be a report that includes all records. For example, take an EREP daily report that processed all records.
   2. Near the top of the report, message IFC120I will tell how many records were written in the time span chosen for the report. Divide this number by the number of seconds in the time span for the average writes per second for this logrec data set.
   3. You can also look at timestamps in the output to analyze how many records were written in a particular second. You can do this to check for peak usage.
   4. Add the results for each logrec data set that will write to the log stream to get the total writes per second for a logrec log stream.