2. PERFORMANCE REPORTING ANALYSIS › 2.4 I/O Configuration Analysis › 2.4.7 Tape Drive Simulation Analysis Inquiry › 2.4.7.2 Technique Tutorial › 2.4.7.2.5 Modeling

2.4.7.2.5 Modeling

The report requests can be accompanied by several other types
of optional exits that expand the straightforward historical
analyses into models.  These models have tape drive
allocations that have been altered in one of several ways
from that actually reported in the CA MICS database.

The reports issued under this modeling feature are identical
to the standard reports, except that each report is clearly
identified as a MODEL report, instead of as an ANALYSIS
report.

The standard modeling features permit two types of
modifications of tape drive allocation:

    o  Specifying additional tape drive use
    o  Moving existing tape use to a different time period


SPECIFYING ADDITIONAL TAPE DRIVE USE

You can specify additional hypothetical use of tape drives
via the Add workload observations extended option.  This
facilitates study of the effects of additional tape drive use
beyond that which already occurs.

MOVING EXISTING TAPE USE TO A DIFFERENT TIME PERIOD

The second feature, moving existing tape use to a different
time period, allows you to specify a new reader timestamp,
RDRTS, for one or more jobs via the Move workload
observations extended option.  The RDRTS represents the date
and time when the job first entered the system.  If one or
more jobs are moved, the new STARTTS and ENDTS are calculated
as follows for each job:

    ENDTS(new) = ENDTS(old) + (RDRTS(new) - RDRTS(old))

    STARTTS(new) = STARTTS(old) + (RDRTS(new) - RDRTS(old))

You can use this feature, for example, to model the shift of
a "tape hog" job to a different time slot.

MODELING CONCERNS

The modeling technique used is very simple.  As a linear
unconstrained model, it does not take into consideration any
resource scarcity or queuing.  Such a model can still be very
helpful when attempting to schedule batch workloads or plan
for hardware requirements.  There are several implications of
using a linear unconstrained model:

   o   Queuing delays due to I/O path contention are not
       included in the model.

   o   There is no initiator or memory queuing.  If a job is
       moved or added, it is assumed by the model that it has
       all other resources necessary to run it, including an
       available initiator and adequate memory.  In addition,
       if one job is moved, the job that actually followed
       it in an initiator would not have its starting time
       automatically altered.

These implications are in general no different from those of
any modeling; you should understand the model's limits and be
sure that the model represents your real world with
sufficient accuracy to be helpful.

More complex modeling, including predicting the effects of
shifting to 3480 drives, can be accomplished using a general
purpose exit that is described in Section 2.4.7.2.6.