Method, system and program product for managing network performance

A service level agreement between a network service provider and a network user may require that specified service metrics or parameters be maintained within predetermined limits. To reduce the chance those limits (and the service level agreement) will be violated, the service metric is sampled periodically over successive sampling intervals. The mean value and the standard deviation of the set of samples obtained during the most recent sampling interval are used to screen out unreliable data. If the set of samples satisfies the reliability screening, the set is used in combination with past acceptable sets to determine the trend in the sampled service metric. If the trend is toward a violation, the time of the violation is predicted based on the assumption the current trend will continue. If the predicted time of violation is sufficiently near the current time, an alert is sent to the service provider to permit the service provider to initiate anticipatory corrective action.