Graphing and interpreting forecast information

You cannot graph forecasts until you accumulate meaningful data for the forecast time period. For example, if you choose one week as the length of the forecasting period (beginning on the current date) and include four periods in the forecast, wait until the end of a month to graph the actual performance data against the forecast.

1. Select Database > Select.

   The Data Source Selection: ODBC dialog box is displayed. For more information, see Selecting-data.

2. From the list of Available Data Sources, select the appropriate summary database and click OK.
3. Select Database > Detail.
   The Data Selection dialog box is displayed.
4. Click in the box for a component to display a selection list of available components and component groups. (If Visualizer supports grouping a component, you will see the Group Name box over the selection list.)
   Select the components to include in the group. In this case you want to compare a forecast record, appended with _F, with a summary record. Press Shift to choose more than one contiguous entry or Ctrl to choose more than one noncontiguous entry.
5. In the Group Name box, type a name for the group.
6. Click Add Group to add the group to the end of the selection list and select the group.
7. Click OK to exit the Data Selection dialog box.
8. From the Graphics menu, choose the graph that corresponds to the forecast values. 
   For example, if you specified a forecast for total CPU utilization, choose that graph.

Strategies for forecasting

Making forecasts requires some advance planning. For example, you need to decide what kinds of information you want to display and how to make it useful. The following are three possible strategies:

• Forecasting an Average Value: Define a single forecast value to represent a simple average. You can then view the number of periods when the system exceeds or falls below your projections. In this case, group together the single forecast and the daily (or weekly) summary for graphing. This graph shows how often your actual measures approach the mean.
• Forecasting Upper and Lower Bounds: Define two separate forecasts; one for a high value and one for a low value. In this case, group together the high and low forecasts as well as the daily (or weekly) summaries for graphing. This graph shows whether your system stayed within the performance bounds.
• Forecasting Minimum, Maximum, and Average Values: Define three separate forecast values: on e for maximum, one for average, and one for minimum. In this case, group the three forecasts with the daily (or weekly) summaries for graphing. This graph shows how to forecast the total performance of your system.