Computer-readable recording medium, failure prediction device and applicability determination method

An applicability determination method performed by a computer, the method comprising:

• generating a failure predictor pattern in accordance with previous cases of failure that has occurred in a first system configuration, the failure predictor pattern being used to detect a predictor of failure in the first system configuration;
• when a system configuration is changed from the first system configuration to a second system configuration, calculating a number of changes that is an accumulated number of times that a change of the configuration item which is included in a system is executed, the number of changes being calculated with the focus on temporal continuity of the system configuration and indicating a feature amount that does not depend on the system configuration;
• calculating a number of differences that is a number of different configuration items between configuration items included in the first system configuration and configuration items included in the second system configuration, the number of differences being calculated with the focus on spatial continuity and indicating a feature amount that does not depend on a change history of the system;
• detecting a predictor of failure by using the generated failure predictor pattern in the second system;
• first determining, when the predictor of failure is detected, whether or not the number of changes is less than a first threshold and whether or not the number of different configuration items is less than a second threshold;
• and
• second determining that the generated failure predictor is valid when the number of changes is less than the first threshold and the number of different configuration items is less than the second threshold and notifying the predictor of the failure to a predetermined destination.

A failure prediction device, comprising:

• a processor configured to execute a process including:
• generating a failure predictor pattern in accordance with previous cases of failure that has occurred in a first system configuration, the failure predictor pattern being used to detect a predictor of failure in the first system configuration;
• when a system configuration is changed from the first system configuration to a second system configuration, calculating a number of changes that is an accumulated number of times that a change of the configuration item which is included in a system is executed, the number of changes being calculated with the focus on temporal continuity of the system configuration and indicating a feature amount that does not depend on the system configuration;
• calculating a number of differences that is a number of different configuration items between configuration items included in the first system configuration and configuration items included in the second system configuration, the number of differences being calculated with the focus on spatial continuity and indicating a feature amount that does not depend on a change history of the system;
• detecting a predictor of failure by using the generated failure predictor pattern in the second system;
• first determining, when the predictor of failure is detected, whether or not the number of changes is less than a first threshold and whether or not the number of different configuration items is less than a second threshold;
• and
• second determining that the generated failure predictor is valid when the number of changes is less than the first threshold and the number of different configuration items is less than the second threshold and notifying the predictor of the failure to a predetermined destination.