Characterizing a computer system using a pattern-recognition model

1. A method for characterizing a computer system using a pattern-recognition model, the method comprising:

in one or more computer systems, performing operations for:

monitoring values for an environmental parameter from a set of sensors associated with the computer system;

calculating a baseline for the environmental parameter based on the monitored values from a subset of the set of sensors;

subtracting the baseline from the monitored values from sensors in the set of sensors to produce compensated values;

using the compensated values as inputs to the pattern-recognition model, which produces estimates for the compensated values based on correlations between the compensated values learned during a training phase;

calculating residuals by subtracting the estimates for the compensated values from the compensated values; and

analyzing the residuals to characterize the computer system.

2. The method of claim 1 , wherein the pattern-recognition model uses a non-linear, non-parametric pattern-recognition technique.

3. The method of claim 1 , wherein the pattern-recognition model uses a multivariate state estimation technique (MSET).

4. The method of claim 1 , wherein the environmental parameter includes at least one of:

a temperature;

a humidity;

a vibration amplitude; and

an electromagnetic signal.

5. The method of claim 1 , wherein the baseline is an average of values from sensors in the set of sensors.

6. The method of claim 5 , wherein the average is calculated from values contemporaneously obtained from sensors in the set of sensors.

7. The method of claim 1 , wherein the training phase includes selecting the subset of sensors used for calculating the baseline.

8. The method of claim 1 , wherein subtracting the baseline from the monitored values from sensors in the set of sensors includes subtracting the baseline from monitored values from sensors in the set of sensors which are different from sensors in the subset of sensors used to calculate the baseline.

9. The method of claim 1 , wherein analyzing the residuals to characterize the computer system includes generating an alarm based on a sequential probability ratio test (SPRT) of the residuals.

10. The method of claim 1 , wherein at least one sensor in the set of sensors associated with the computer system is located in a separate computer system.

11. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for characterizing a computer system using a pattern-recognition model, the method comprising:

monitoring values for an environmental parameter from a set of sensors associated with the computer system;

calculating a baseline for the environmental parameter based on the monitored values from a subset of the set of sensors;

subtracting the baseline from the monitored values from sensors in the set of sensors to produce compensated values;

using the compensated values as inputs to the pattern-recognition model, which produces estimates for the compensated values based on correlations between the compensated values learned during a training phase;

calculating residuals by subtracting the estimates for the compensated values from the compensated values; and

analyzing the residuals to characterize the computer system.

12. The computer-readable storage medium of claim 11 , wherein the pattern-recognition model uses a non-linear, non-parametric pattern-recognition technique.

13. The computer-readable storage medium of claim 11 , wherein the pattern-recognition model uses a multivariate state estimation technique (MSET).

14. The computer-readable storage medium of claim 11 , wherein the environmental parameter includes at least one of:

a temperature;

a humidity;

a vibration amplitude; and

an electromagnetic signal.

15. The computer-readable storage medium of claim 11 , wherein the baseline is an average of values from sensors in the set of sensors.

16. The computer-readable storage medium of claim 15 , wherein the average is calculated from values contemporaneously obtained from sensors in the set of sensors.

17. The computer-readable storage medium of claim 11 , wherein the training phase includes selecting the subset of sensors used for calculating the baseline.

18. The computer-readable storage medium of claim 11 , wherein subtracting the baseline from the monitored values from sensors in the set of sensors includes subtracting the baseline from monitored values from sensors in the set of sensors which are different from sensors in the subset of sensors used to calculate the baseline.

19. The computer-readable storage medium of claim 11 , wherein analyzing the residuals to characterize the computer system includes generating an alarm based on a sequential probability ratio test (SPRT) of the residuals.

20. An apparatus for characterizing a computer system using a pattern-recognition model, the apparatus comprising:

a monitoring mechanism configured to monitor values for an environmental parameter from a set of sensors associated with the computer system;

a calculating mechanism configured to calculate a baseline for the environmental parameter based on the monitored values from a subset of the set of sensors;

a subtracting mechanism configured to subtract the baseline from the monitored values from sensors in the set of sensors to produce compensated values;

a mechanism configured to use the compensated values as inputs to the pattern-recognition model, which produces estimates for the compensated values based on correlations between the compensated values learned during a training phase;

a residual calculating mechanism configured to calculate residuals by subtracting the estimates for the compensated values from the compensated values; and

an analyzing mechanism configured to analyze the residuals to characterize the computer system.