2. SAMPLE PLANNING APPLICATIONS › 2.15 Tandem CPU Resource Planning › 2.15.4 Data Elements Overview

2.15.4 Data Elements Overview

 
The intent of this section is to provide the capacity planner
with an understanding of the information chosen for the
Tandem CPU Resource Sample Application and how these
variables can be applied to capacity planning situations in a
Tandem environment.

CPETYPE - Tandem CPU Type
 
The CPU type of the Tandem processor on which MEASURE data
was recorded.  It is alphanumeric.  Although not directly
useful for capacity planning, CPETYPE provides an audit trail
of changes that may have occurred in your Tandem systems.

CPECBSTM - CPU Busy Time
 
The amount of time that the CPU was executing instructions.
CPECBSTM is maintained as a TIME14.2 SAS variable.  Unlike
Percent CPU Busy (CPEPCBSY), this is a direct measurement of
the amount of time that the CPU was active.  Because CPU is
the most critical resource, and CPECBSTM determines the
amount being used, this is one of the most important
variables in capacity planning, regardless of the operating
environment.

The capacity planner should watch this value closely and be
aware of any unusual fluctuations in its size (either up or
down).  This variable can be used as a forecast element in
univariate modeling, however, be wary of changes in Tandem
CPU Type (CPETYPE) that may affect the rate at which
instructions are executed.

CPEPCBSY - Percent CPU Busy
 
A numeric variable that reflects the percentage of uptime
that the CPU was busy executing instructions.  The value
ranges from 0 to 100% (see Section 2.15.3) for the
derivation) and is a relative indicator of how much CPU
resource is consumed (or remaining).  Like CPU Busy Time
(CPECBSTM), this value can be used to compare the activity of
one system (or CPU) to another.

CPEPCBSY provides a simple indicator of processor resources
and can also be used for forecasting purposes.  Although it
provides the same information as CPECBSTM, it is more helpful
because you do not need to examine uptime to see how much CPU
time is available.  It is also useful for presenting a
capacity plan to management.

CPEPCBSY is the most commonly used value for capacity
planning.  Other elements from the CPU file are included in
the capacity database file because they may be of some
interest to the capacity planner, however, CPEPCBSY is the
recommended element for forecasting CPU capacity needs.

CPEDIOS - Disk I/Os
 
The number of disk I/Os processed in the CPU.  It is an
accumulated numeric variable and is limited only by practical
bounds.  CPEDIOS is a relative indicator of I/O activity on
the system.

I/O is second only to CPU utilization when evaluating the
capacity of a computer system.  Increases in I/O indicate
growth in workloads on the system.  Use CPEDIOS to forecast
when the I/O counts may cause system saturation.

CPEPSDIO - Disk I/Os Per Second
 
The rate of disk I/Os per second of uptime for the processor.
See Section 2.15.3 for its derivation.  It is numeric and
limited only by practical bounds.  Like CPEDIOS, CPEPSDIO is
a relative indicator of I/O activity on the system.

Because CPEPSDIO is in rate per second rather than total
count, it is more suitable for trending and forecasting than
CPEDIOS.  Monitor the trend of I/Os per second to determine
when your system will reach its I/O processing saturation
limit.

CPEDISP - Number of Process Dispatches
 
The number of processes dispatched by the CPU.  It is an
accumulated numeric variable and is limited only by practical
bounds.  The number of dispatches is an indicator of the
amount of work being completed on the system.
 
Temporary increases in CPEDISP indicate periods of heavy work
on the system, while a constant increase in this value
indicates steady growth in workloads.

CPEIBSTM - Interrupt Busy Time
 
The amount of time the CPU was busy handling interrupts.
CPEIBSTM is maintained as a time variable.  The interrupts
are usually for processing of messages and I/O operations.
Monitor the changes CPEIBSTM in comparison to I/O rates and
messages sent or received for indications of performance
problems.

Disproportional increases in CPEIBSTM to the total CPU Busy
Time (CPECBSTM) may indicate problems with I/O or message
processes.  You may want to report CPEIBSTM as a percentage
of CPECBSTM.

CPEOVHTM - Process Overhead Time
 
The CPU time spent on overhead that cannot be identified with
a particular process.  It is maintained as a time variable.
It is limited only by practical bounds, however, it is
usually a small percentage of the total CPU Busy Time
(CPECBSTM).  Like CPEIBSTM, you may want to report CPEOVHTM
as a percentage of CPECBSTM.

CPEPCOBS - Percent CPU Process Overhead Busy
 
The percentage of uptime that the CPU was busy with overhead
processing.  See Section 2.15.3 for information on its
derivation.  The value of CPEPCOBS ranges from 0 to 100%,
however, it is usually only a fraction of a percent of the
total processor uptime.

CPEPCOBS is most useful when shown as a percentage of the
total CPU Busy Time (CPECBSTM).  Low overhead busy time
indicates efficient use of the processor.  You may want to
use this element to measure the trend of overhead busy time
over an extended period of time. Disproportional increases in
the element should be cause for concern.

CPEPSSTM - Send Busy Time
 
The amount of time the CPU was busy sending data to another
system.  CPEPSSTM is maintained as a time variable and is
limited only by practical bounds.  There may be a slight
overlap in CPEPSSTM and Interrupt Busy Time (CPEIBSTM).

Like CPEIBSTM, CPEPSSTM is most useful when shown as a
percentage of Total CPU Busy (CPECBSTM).  You may want to
look at this element for trends over a long period of time.

CPEPCSBS - Percent CPU Send Busy
 
The percentage of uptime the processor is busy sending data
to another system.  It is maintained as a numeric variable
and ranges from 0 to 100%.  See Section 2.15.3 for the
derivation.  Like other percent variables, you can determine
how much of the available time is being used without
examining other variables.

CPEPCSBS is most meaningful when compared to the total CPU
Busy Time (CPECBSTM).  Monitor the trend of this element to
see what percentage of processor time is used for interaction
between systems.

CPEPCIBS - Percent CPU Interrupt Busy
 
The percentage of uptime the processor is busy handling
system interrupts.  CPEPCIBS is maintained as a numeric
element and ranges from 0 to 100%.  See Section 2.15.3 for
the derivation.  Like other percent variables, you can
determine how much of the available time is being used
without examining other variables.

CPEPCIBS is most meaningful when compared to the total CPU
Busy Time (CPECBSTM).  Monitor the trend of this element to
see what percentage of processor time is used for handling
interrupts.

CPECAHIT - Number of Cache Hits
 
The number of times the required data was found in cache and
resulted in a reduced disk I/O count.  It is numeric and
limited only by practical bounds.
 
The capacity planner should monitor this value.  Decreased
cache hits may indicate a need for more cache.

CPEPSCAH - Cache Hits Per Second
 
CPEPSCAH measures the rate of cache hits per second and is
numeric.  See Section 2.15.3 for the derivation.  Higher
rates indicate better performance and a sufficient capacity
of cache resources.
 
The capacity planner should watch for decreases in this rate
as an indication the system is approaching its capacity
limit.  Decreases in the cache hits per second will result in
increased disk I/Os.

CPEPSDSP - Dispatches Per Second
 
The rate of dispatches per second for the processor.  It is
numeric and is limited only by practical bounds.  See Section
2.15.3 of this guide for the derivation.  Because CPEPSDSP
factors in the uptime, this number is more useful for
forecasting dispatches.

The capacity planner will use this number to determine the
throughput rate of the processor.  An increase in the
dispatch rate may indicate an increase in the work being done
or a change in the mix of work.  When the dispatch rate is a
constant value, review system use carefully to determine
whether the rate is constant because all necessary work is
being completed or because the system is saturated.

CPEPSSWP - Swaps Per Second
 
The swap rate per second for the processor.  It is a numeric
value and is limited only by practical bounds.  See Section
2.15.3 for the derivation.
 
CPEPSSWP is an indicator of memory usage.  Low rates indicate
adequate memory, while higher rates indicate contention for
the available memory on the system.  Monitor this value to
forecast memory requirements.

CPESWAPS - CPU Swaps
 
CPESWAPS is an accumulated numeric counter of swaps in and
out of the system by the memory manager.  It is limited only
by practical bounds.  The swap count is a count of the number
of swaps, not the number of pages swapped.
 
The capacity planner will monitor this value to determine
when more memory is required.  Because this is a counter and
not a rate, be aware of the interval duration when using this
variable.

DURATION - Recording Interval
 
DURATION is a time variable indicating the duration for the
interval used for this observation in the capacity database
file.  The duration value is taken from the TDMCPE file and
may be less than the system duration, SYSDUR, if the
processor was not active for the full interval.  DURATION is
used in rate and utilization calculations when the CAPAPU
value is CPUNUM.

This value is included on the capacity database file for
informational purposes.  It may serve as an indicator of
missing processor data.
 
Note:  If Tandem System Category (CAPAPU) is NONE or EXIT,
       the System Profile File (TDMSYP) is merged with the
       CPU Entity File (TDMCPE) to determine the duration of
       the intervals for the complete system.  Only in this
       circumstance are the remaining elements kept on the
       capacity database file.

SYPNOCPU - Maximum Number of CPU Processors
 
The maximum number of CPU processors in use for the entire
system during this interval.  It is numeric from 1 to 16.
 
SYPNOCPU is kept on the capacity database file for
informational purposes and is not recommended as a forecast
element.

SYPTOUTM - Total System Uptime
 
The total time the system is active.  SYPTOUTM is stored as a
time element.  It is derived, as shown in Section 2.15.3, by
the Tandem CPU Resource Sample Application.

SYPTOUTM is used as the denominator for calculating rates and
percent utilization when data is summarized across CPU
processors.

SYPTOUTM is kept in the capacity database file because it is
used to derive other elements.  It is not recommended as a
forecast element.

CPUCOUNT - Number of CPU Processors Found in the TDMCPE File
 
The number of CPU processors in the TDMCPE file.  It is
numeric and accumulated.  Because data is summarized across
CPU processors when NONE or EXIT is specified, you need to
know how many CPU processors are included in the system or
subsystem represented in the capacity planning database file.
CPUCOUNT is used in the derivation of SYSTOUTM.

CPUCOUNT is included in the capacity database file because
it is used to derive other values.  It is not recommended as
a forecasting element.

SYSDUR - System Duration or System Uptime
 
The duration or uptime of the entire system for an interval.
It is stored as a time element.  The derivation is shown in
Section 2.15.3.  SYSDUR is used when CAPAPU is EXIT or NONE,
rather than using the duration from the individual processors
in the TDMCPE file, because it gives the duration of the
entire system.  For example, for the WEEKS timespan, for hour
10, for the prime shift (Monday through Friday), SYSDUR would
be 5 hours.