14. WORKLOAD FORECASTING › 14.2 Usage Guidelines › 14.2.4 Forecasting Future Values with the Model

14.2.4 Forecasting Future Values with the Model

The final step in developing the Workload Forecasting
Analysis model is to use the model to predict future values
of the Forecast resource variable.  This is easily
accomplished using the Workload Forecasting Analysis model
along with Neugents technology.

Predicting future values of the Forecast element value
requires the analyst to first extrapolate the values of the
selected Analysis element values and then to use those values
as input to the model prediction process.  Workload
Forecasting Analysis provides assistance in this process with
internal options for extrapolating the Analysis data values,
so that you do not need to rely on external means.  You may
choose to adjust the Analysis data values using either the
MANUAL option, where an adjustment value (as a percentage) is
applied against each selected Analysis element (workload), or
the AUTOMATIC option where the application examines the
Analysis element values to determine the growth rate of each
workload, and then assigns each respective growth rate
internally.  Once the growth rate has been assigned, the
Workload Forecasting Analysis application extrapolates the
data values for each selected Analysis element for the number
of cycles specified in the Forecast Length. Neugents
technology is then invoked to predict future values of the
Forecast variable using the newly extrapolated data values as
input. The output of the forecasting execution is a file
containing the original input values, including the Forecast
and selected Analysis element values, and all output values.
This file is used to generate a variety of reports and
graphs.

One consideration in any forecasting operation is "where to
start".  That is to say, what point in time is appropriate to
use as a starting data value for the forecasting operation.
Conventional wisdom says to use the last known set of
Analysis element values (workloads) to begin forecasting, but
this can result in misleading results.  Suppose, for example
that the most recent set of data points occurred in a time
period of reduced usage. Starting forecasting from that point
might well result in a forecast that is too low in magnitude.
The opposite could certainly occur if the starting set of
data points were higher than what normally occurred in
operation of the system.  To address that issue, the Workload
Forecasting Analysis application first reviews the historical
data values for each of the Analysis elements independently,
determines an adjustment factor for each value and creates a
starting set of data points from the adjusted set. Note that
while this set of data points may well mimic a real point in
time, it is only a base line estimate but will generally
provide better results when the input history data is not
well behaved.

The newly predicted values of the Forecast element, along
with the extrapolated values of the Analysis elements can be
optionally saved in a CA MICS Capacity Planner Forecast file
for additional analysis or presentation purposes.