Trending Method Option [ HP RXForecast Users Manual for MPE Systems ] MPE/iX 5.0 Documentation
HP RXForecast Users Manual for MPE Systems
Trending Method Option
The Trending Method option appears on the Forecasts command dialog box.
Use this option to determine which trending method HP RXForecast will use
to model. Choose the method based on the way you expect your data to
change over time.
Smoothing
The Smoothing method does not forecast at all. Instead, it displays the
trend over time. The method itself is called median smoothing. It
replaces the values with medians of surrounding values until a smooth
curve is obtained. No Interval Types option is available with this
technique.
Linear
The Linear method is the most common and simplest method for trending.
It models the trend as a straight line.
Exponential
The Exponential method is similar to the Linear method. The trend,
however, is projected using an exponentially increasing function. This
may be applicable in the case of response times that tend to increase
slowly when the load is light, but quickly when the load is heavy.
S-Curve
The S-Curve method is another function that can be used in certain
situations. If the metric has the same behavior as the exponential
function except that it levels off at some value, this method may be
useful. For example, CPU Utilization may start off stable at a certain
value, but increase rapidly until about 90 percent and then level off;
slowly approaching, but never exceeding, 100 percent.
Business Units
In the Business Units method, a trend is correlated to a known business
unit that the user constructs.
Two kinds of information are necessary to build a business unit:
* Historic data on the business unit.
* Future values forecasted for the business unit.
In order to forecast past the last date in the logfile, you must
forecast the business unit into the future and include those
values in the business unit file. Typically, this forecast would
be derived from a business plan or projections made by a planning
group.
In short, you need to know where the business unit has been and where it
is likely to go.
Once you have the necessary information on the business unit, you can
construct a business unit file by entering the business units and
applicable dates into a flat ASCII file. See "Creating a Business Unit
File" .
In order to use the Business Units trending method, you must correlate a
trend to a known, user-defined business unit. Selection of a business
unit should be based on factors that affect the specific computer system.
A good business unit has strong correlation with the metric being
forecast. The Business Units trending method is most successful when
such a correlation exists.
The Business Units trending method is dimmed (cannot be selected) until a
business unit file has been selected on the NEXT screen.
MPE/iX 5.0 Documentation