1 02/28/85  work_class_meters, wcm
  2 
  3 Syntax as a command:  wcm {-control_arg}
  4 
  5 
  6 Function:  prints certain information from the tc_data segment about
  7 each work class currently defined.
  8 
  9 
 10 Control arguments:
 11 -report_reset, -rr
 12    generates a full report and then performs the reset operation.
 13 -reset, -rs
 14    resets the metering interval for the invoking process so that the
 15    interval begins at the last call with -reset specified.  If -reset
 16    has never been given in a process, it is equivalent to having been
 17    specified at system initialization time.
 18 
 19 
 20 Access required:  This command requires access to phcs_ or
 21 metering_gate_.  Additionally, in order for the command to print the
 22 names of the work classes, access to both the Master Group Table (MGT)
 23 and the answer table is required.  These tables are located in the
 24 directory >system_control_1.
 25 
 26 
 27 Notes:  If the work_class_meters command is given with no control
 28 argument, it prints a full report.
 29 
 30 When the scheduler is operating in percent mode, percentages are
 31 computed against two different base CPU quantities.  It is necessary to
 32 understand the differences between these quantities in order to
 33 interpret the output of work_class_meters.
 34 
 35 
 36 One base quantity is the total system CPU time.  This is simply the
 37 total realtime all CPUs have been active doing anything (including
 38 running an idle process).  In any interval of time when there was no
 39 reconfiguration of CPUs, the total system CPU time is the product of
 40 the length of the interval and the number of CPUs.  Another base
 41 quantity is nonidle CPU time.  This is the total CPU time expended by
 42 all CPUs except when running an idle process.  It is given by the total
 43 system CPU time minus the sum of all idle time reported by
 44 total_time_meters (MP Idle, Non-MP Idle, and Zero Idle).
 45 
 46 
 47 When the scheduler is operating in percent mode, it distributes CPU
 48 resources among contending work classes according to their guaranteed
 49 percentages.  These percentages are percentages of total nonidle CPU
 50 time.  So, if there are two work classes, each with a guarantee of 50
 51 percent, and the system is 50 percent idle, each work class gets 25
 52 percent of total system CPU time (assuming that there is enough demand
 53 for this to be possible).  In this example, each work class is getting
 54 50 percent of the nonidle CPU time, but only 25 percent of the total
 55 system CPU time.  Another way of viewing this is that the guaranteed
 56 percentages define a relationship among work classes according to the
 57 ratio of percentages.  That is, a work class with a guaranteed
 58 percentage of 10 percent gets about half as much CPU time as a work
 59 class with a guaranteed percentage of 20 percent, assuming sufficient
 60 demand by both.  Further, this ratio is independent of the system load.
 61 
 62 
 63 The system administrator can limit the CPU resources consumed by a work
 64 class to a fixed percentage of the total system CPU time.  The
 65 scheduler enforces this limitation, even at the expense of going idle.
 66 That is, a work class with a maximum percentage of 10 percent gets no
 67 more than 10 percent of the total CPU time in any interval, regardless
 68 of load.  Excess CPU time is distributed among work classes with no
 69 maximum percentage, according to their guaranteed percentages.  If this
 70 cannot be done, the excess CPU time becomes idle time.
 71 
 72 At any time one or more work classes may be a realtime work class with
 73 specified response time and quanta.  A process in such a work class is
 74 low priority until its deadline arrives, at which time it is made
 75 eligible regardless of any other constraints.  The remainder of the
 76 work classes are scheduled either by percentage of CPU time (percentage
 77 mode) or by soft deadlines (deadline mode).
 78 
 79 
 80 The following parameters are always displayed for each work class.
 81 WC
 82    is the number of the work class.
 83 %GUAR
 84    is the percentage of nonidle CPU time guaranteed to the work class
 85    if the scheduler is being operated in percent mode and if there is
 86    sufficient demand by the work class for this to be possible.
 87 %MAX
 88    is the maximum percentage of total CPU time allowed by the system
 89    administrator to be consumed by this work class.  This field is
 90    blank if the work class has no limitation on CPU consumption.
 91 
 92 
 93 %TCPU
 94    is the percentage of total CPU time actually received by this work
 95    class in the metering interval.
 96 V/ELIG
 97    is the average amount of CPU time used per eligibility quantum.
 98 PW
 99    is the pin weight, or number of free laps for pages brought into
100    memory.
101 
102 The following parameters are always displayed for realtime work
103 classes, and are displayed for other work classes only if the scheduler
104 is operating in deadline mode.
105 IRESP
106    is the response time (in seconds) specified for the work class after
107    an interaction.
108 
109 
110 IQUANT
111    is the initial quantum (in seconds) for the work class after an
112    interaction.
113 RESP
114    is the specified delay (in seconds) between subsequent quanta.
115 QUANT
116    is the value (in seconds) of subsequent quanta.
117 
118 The following parameters are displayed when the scheduler is operating
119 in either deadline or percentage mode.
120 P
121    if printed, members of the work class are post purged.
122 
123 
124 M
125    is the max_eligible limit per work class.  A zero means the work
126    class has no particular limit.
127 R2
128    if printed, the members of the work class are scheduled in realtime
129    mode.  They are made eligible at or before their deadlines.
130 I
131    if printed, members of the work class are given scheduler priority
132    after interactions.
133 LCG
134    are the load control groups that are placed in the work class.  If
135    the LCG name is parenthesized, only the absentee processes in the
136    LCG are placed in the work class.