6. DATA SOURCES › 6.8 PR/SM LPAR Concepts › 6.8.5 Data Analysis › 6.8.5.2 LPAR Level Analysis › 6.8.5.2.1 General Purpose LPA/Rs › 6.8.5.2.1.1 z/OS LPARs Without Specialized Processors

6.8.5.2.1.1 z/OS LPARs Without Specialized Processors

 
 If a z/OS logical partition does not use any specialized
 processors such as zAAP, zIIP, ICF or IFL then the only type
 of CPU resources it can consume comes from the standard CP
 pool of processors of the CPC.  Your ability to accurately
 track its utilization depends on two key factors:  does it
 use shared or dedicated processors, and is the wait assist
 facility enabled?
 
 
 Shared Standard CP Processors
 -----------------------------
 
 If the LPAR is configured with shared standard CP processors
 and wait assist is disabled (Wait Completion = No), the time
 the logical processors are dispatched, from PR/SM management
 routines, is equal to the actual busy time as seen from
 inside the LPAR.  In other words, the value of LPCTODTM in
 the HARLPC file is the same as CPUTODTM as well as CPUTOBTM
 in the HARCPU file.
 
 Although, in theory, you could choose to report from either
 the HARCPU or the HARLPC files, we recommend you use the
 HARLPC file measurements.  The main reason for this is that
 the HARCPU file will only contain information for logical
 partitions running RMF and whose SMF type 70 subtype 1
 records are processed by CA MICS.  For example, the HARCPU
 file should not be used for non z/OS LPARs, or if you input
 to CA MICS the SMF records from only one of your z/OS LPARs.
 This problem does not apply to the HARLPC file because, as
 long as one or more logical partitions has its SMF type 70
 subtype 1 records processed by CA MICS, it will contain
 information about all LPARs defined in the configuration.
 
 However, care must be taken when using the HARLPC file
 because, if multiple z/OS systems are active in the complex,
 each one can record the activity of the others.  It is
 therefore necessary to select only one system (SYSID) from
 the HARLPC file, in order to avoid data duplication.  We
 recommend that you select the SYSID that is most often active
 during the day.  If you have 3 LPARs defined and one is a
 production machine that will be active all the time, and the
 other 2 are test machines that come up and down, use the
 production machine's SYSID.
 
 Listed below are examples for your shared LPARs analysis:
 
 o To know how much of its allocated CPU resources an LPAR
   consumed, use the following data element from the HARLPC
   file:
 
   LPCPCVPU - LPAR CP Processors Utilization
 
   LPCPCVPU is computed by the formula:
 
     LPCTODTM*100 / LPCTOOTM
 
   where LPCTODTM is the Total CP Processors Dispatch Time and
   LPCTOOTM is the Total CP Processors Online Time.
 
 o To know how many of the resources of the pool of standard
   CP processors were used by one LPAR, use the following data
   element from the HARLPC file:
 
   LPCPCTSU - LPAR Total CP Processors Utilization
 
   LPCPCTSU is computed by the formula:
 
     100 * LPCTODTM
     --------------------
     DURATION * PRSMTCP
                --------
                INTERVLS
 
   where DURATION contains the number of seconds in the
   recording interval, PRSMTCP is the Total CP Processors in
   the Complex, and INTERVLS is the number of RMF intervals.
 
   Note:  If all processors from the standard CP pool are
   shared, this element is equal to LPCPCSSU (LPAR Shared CP
   Processors Utilization).
 
 
 A Capacity Planning Consideration
 ---------------------------------
 
 When using either LPCPCVPU or LPCPCTSU in capacity planning
 studies, keep in mind that LPCPCTSU is the percent of CP
 processors utilization over all of the CP processors in the
 complex.  This includes processors that are not active.
 
 Since LPCPCVPU is the CP processors utilization over only
 those processors that are online, you should decide whether
 LPCPCVPU or LPCPCTSU is more appropriate for your capacity
 analysis.
 
 
 Dedicated Standard CP Processors
 --------------------------------
 
 When the logical processors of an LPAR configured with
 dedicated standard CP processors are dispatched, the
 resources they use are unavailable for other use whether they
 are actually executing instructions or are in the wait state.
 From the PR/SM perspective, the dispatch time is equal to the
 measurement interval time, regardless of actual processor
 utilization, and the LPAR appears as being used 100% of the
 time in the HARLPC file.
 
 In order to determine the actual utilization, the total time
 in seconds that the logical processors were not busy (wait
 time) must be subtracted from the total dispatch time.
 Since this wait time is not available in the HARLPC file,
 dedicated LPARs' utilization must be tracked from the HARCPU
 file.  However, in regards to the amount of total CPU
 resources allocated to these LPARs, and thus unavailable to
 others, you can still use information from the HARLPC file.
 
 Listed below are examples for your dedicated LPARs analysis:
 
 o To know how much of its allocated CPU resources an LPAR
   consumed, use the following data element from the HARCPU
   file:
 
   CPUPCBSY - Pct CP Processors Busy
 
 o To know how many of the resources of the pool of standard
   CP processors were dispatched to one LPAR, use the
   following data element from the HARLPC file:
 
   LPCPCTSU - LPAR Total CP Processors Utilization
 
   Note:  If there are dedicated processors that are not
   active, then they would not be counted among the online
   processors during an RMF interval.  In this case, using
   LPCPCVPU might be more appropriate for capacity studies.
 
 Wait Assist Enabled
 -------------------
 
 When the Wait Assist function is enabled, it indicates that
 the LPAR's logical processors remain dispatched until the end
 of the user-specified interval, called time slice, even when
 they are in the wait state. In this case dispatch time
 includes wait time.
 
 Dedicated LPARs do not use the wait-assist feature because
 they are dedicated to specific physical processors and have
 them whether or not they wait.
 
 Shared LPARs are where the wait-assist feature comes into
 play.  As for dedicated LPARs, if the feature is enabled,
 although you can still use the information from the HARLPC
 file to get the amount of total CPU resources allocated to
 such an LPAR, you must look into the HARCPU file to get the
 actual LPAR utilization.
 
 o To know how much of its allocated CPU resources an LPAR
   consumed, use the following data element from the HARCPU
   file:
 
   CPUPCBSY - Pct CP Processors Busy
 
 o To know how many of the resources of the pool of standard
   CP processors were dispatched to an LPAR, use the following
   data element from the HARLPC file:
 
   LPCPCTSU - LPAR Total CP Processors Utilization
 
   Note:  If you need to know the percent of the pool of
   online standard CP processors were dispatched to an LPAR,
   use the LPCPCVPU element.