3. CAPACITY PLANNING DATABASE › 3.3 Resource Element Files › 3.3.3 Resource Element File Case Studies › 3.3.3.1 Extended Example I/O Workload Analysis › 3.3.3.1.1 Problem Description

3.3.3.1.1 Problem Description

 
You can use the capacity planning resource element file to
collect monthly historical data to analyze the growth in the
utilization of the base group of selector and block
multiplexor channels on an IBM 9021-711. The scope of this
study is to determine if load growth requires the acquisition
of the optional channel group.

The base channel group of a 9021-711 consists of one-byte
multiplexor channel (channel 0) and five block multiplexor
channels (channels 1 though 5).  Therefore, data is collected
for channels 1 through 5 of the system.  Two values
characterize the workload of each of these channels:
 
o  The number of successful start I/Os (SIOs) issued by the
   channel.  This value is provided by the CA MICS data
   element PCASIO in the HARPCA File.
 
o  The utilization (that is, percent busy) of the channel.
   This value is provided by the CA MICS data element
   PCAPCBSY in the HARPCA File.
The following is a discussion of the design of the control
stream that generates these values.

CONTROL STREAM DESIGN
 
When you select a CA MICS file, such as the HARPCA file, as
input to the resource element files, always consult the file
description in the CA MICS Database Structure and Content
Guide to ensure that the required data elements are not lost
when the file is sorted and summarized by the default
sequence that is used by the capacity planning database
generator.  This sequence is
 
    SYSID ZONE YEAR timespan

For the HARPCA File, the summary sequence for the MONTHS
timespan is:
 
    SYSID CHANNEL CPUID YEAR MONTH ZONE

If the HARPCA file is summarized by the default sequence, the
data for the individual channels is summarized into one
meaningless observation unless you add CHANNEL to the summary
sequence of the file.  Since the system is a uniprocessor,
the CPUID is unimportant.

The program initially selects the data elements that you
specify from the CA MICS file. In the data selection step,
the program invokes the user-defined selection exits. These
exit macros perform two functions:
 
o  Deleting unnecessary observations to reduce the processing
   requirements of the program
 
o  Calculating user-defined data elements

For this extended example, the selection exit is used to
delete the observations for the channels that are not
involved in the study.  While in this example deleting the
unnecessary channel observations represents only a small
potential savings, the exit macros are introduced to remind
you of the potential resource savings when files like the
BATJOB or BATPGM files are processed.

After the HARPCA data is summarized, one record for each
unique SYSID ZONE YEAR MONTH CHANNEL sequence exists in the
file.  This data set is processed by a data step that invokes
the user-supplied summary logic that is coded in the summary
exit macro.  The user-supplied summary logic is used to
assign the PCASIO and PCAPCBSY values to the user-defined
data elements for the individual channels.

For this example, the user-defined data elements are the
following:
 
    CHAN1BS & CHAN1CNT - the  utilization and SIO values for
                         channel 1
    CHAN2BS & CHAN2CNT - the  utilization and SIO values for
                         channel 2
    CHAN3BS & CHAN3CNT - the  utilization and SIO values for
                         channel 3
    CHAN4BS & CHAN4CNT - the  utilization and SIO values for
                         channel 4
    CHAN5BS & CHAN5CNT - the  utilization and SIO values for
                         channel 5

These values are assigned by testing the value of the CHANNEL
data element and then assigning the PCASIO and PCAPCBSY
values to the user-defined data elements.
 
After the data step in which the user summary exit is
invoked, the observations are sorted into the default summary
sequence for the final merge step.  This summary sequence is
 
    SYSID ZONE YEAR timespan

The records are then merged with the default sequence key.
After the records are merged, the program drops the summary
and drop variables, resulting in a single record produced for
each month that contains the five channel utilizations
(CHAN1BS through CHAN5BS) and the five SIO counts (CHAN1CNT
through CHAN5CNT).