Linear Regression

Averages and maximums are based on either daily raw data or hourly raw data, depending on the global time context. Hourly aggregates are used when the time period is within the last seven days, and daily aggregates are used otherwise.

Forecast periods are set to one-half the length of the trending time period. For example, 30-day trends are projected 15 days forward, and 90-day trends are projected 45 days forward.

The following chart shows average and maximum CPU usage for the SALES-SP database using the Linear regression forecast model.

The forecasted maximum and average usage both rise above the current allocation line. This suggests that additional capacity might be required if the current growth trend continues.