Use of operational configuration parameters to predict system failures

1. A method, comprising:

initializing a digital system, the initializing comprising determining an operational configuration of at least part of the digital system;

saving the operational configuration to a database stored on the digital system;

reading the operational configuration from the database and comparing the operational configuration to a reference configuration; and

identifying the digital system as being at risk of a future failure if at least one parameter of the operational configuration differs from the at least one same parameter of the reference configuration by more than a tolerance value;

wherein determining the operational configuration comprises:

a controller within the digital system writing training sequence data to a device;

the controller performing one or more reads from the device of the training sequence data, the controller configured using a distinct configuration of a plurality of configurations for each of the one or more reads; and

the controller selecting the operational configuration from one of the plurality of configurations associated with a successful read.

2. The method of claim 1 , further comprising:

collecting a plurality of operational configurations each corresponding to a distinct digital system of a plurality of digital systems; and

defining each parameter of the reference configuration as a statistical combination of the corresponding parameters from each of the plurality of configurations.

3. The method of claim 2 , wherein defining each parameter of the reference configuration comprises calculating the mean value of the corresponding parameters.

4. The method of claim 2 , further comprising defining the tolerance value as a multiple of a standard deviation referenced to the statistical combination of the corresponding parameters.

5. The method of claim 2 , further comprising:

identifying, as statistical outliers, values of a parameter from a subset of the plurality of operational configurations that each differ by more than the tolerance value from a calculated statistical norm associated with the same parameter for the plurality of operational parameters;

correlating the statistical outliers to actual system failures; and

selecting, as the at least one same parameter, one or more parameters from the subset comprising parameters with correlations to the actual system failures higher than correlations of other remaining parameters from the subset.

6. The method of claim 2 , further comprising:

identifying, as statistical outliers, changes over time in the values of a parameter from a subset of the plurality of operational configurations that each differ by more than the tolerance value from a calculated statistical norm associated with changes over time in the same parameter for the plurality of operational parameters;

correlating the statistical outliers to actual system failures; and

selecting, as the at least one same parameter, one or more parameters from the subset comprising correlations to the actual system failures higher than correlations of other remaining parameters from the subset.

7. A system, comprising:

processing logic; and

a non-volatile storage device coupled to the processor;

wherein the processing logic initializes a configuration of the system and stores at least part of the configuration in a database stored on the non-volatile storage device; and

wherein the processing logic retrieves and makes available stored configuration data for comparison with at least part of a reference configuration to determine if the system is at risk of a future failure;

wherein the system further comprises advanced configuration and power interface (ACPI) registers associated with restoring the system state upon exiting the ACPI global sleep state (ACPI S3 Restore Registers), and wherein the stored configuration comprises values stored in the ACPI S3 Restore Registers.

8. The system of claim 7 , further comprising:

a memory subsystem coupled to the processing logic and comprising volatile memory;

a first bridge coupled to the processing logic, to the volatile memory, and to a first bus; and

a second bridge coupled to the first bridge and to a second bus coupled to the non-volatile storage device.

9. The system of claim 8 , wherein the first bridge comprises a memory controller hub (MCH), and wherein the stored configuration data comprises values stored in registers within the MCH.

10. The system of claim 8 , wherein the memory subsystem further comprises an advanced memory buffer (AMB) chip, and wherein the stored configuration data comprises values stored in registers within the AMB chip.

11. The system of claim 8 , wherein the first bridge comprises a peripheral component interconnect express (PCIe) interface, and wherein the stored configuration data comprises values stored in registers within the PCIe interface.

12. The system of claim 7 wherein the database comprises an extended markup language (XML) database.

13. A computer-readable medium comprising software that causes a first processor to:

cause a controller within the system to write training sequence data to a device;

cause the controller to perform one or more reads from the device of the training sequence data, the controller configured using a distinct configuration of a plurality of configurations for each of the one or more reads; and

cause the controller to select the operational configuration from one of the plurality of configurations associated with a successful read compare the selected operational configuration to a reference configuration, wherein said operational configuration includes at least one parameter that is configured during initialization of said system; and

identify the system as being at risk of a future failure if at least one parameter of the operational configuration differs from the at least one same parameter of the reference configuration by more than a tolerance value.

14. The computer-readable medium of claim 13 , wherein the software further causes a second processor to:

save the selected operational configuration to a database stored on the system.

15. The computer-readable medium of claim 14 , wherein the software further causes the second processor to:

collect a plurality of operational configurations each corresponding to a distinct system of a plurality of systems; and

define each parameter of the reference configuration as a statistical combination of the corresponding parameters from each of the plurality of configurations.

16. The computer-readable medium of claim 15 , wherein the software causes the second processor to define each parameter of the reference configuration by at least calculating the mean value of the corresponding parameters.

17. The computer-readable medium of claim 15 , wherein the software further causes the second processor to define the tolerance value as a multiple of a standard deviation referenced to the statistical combination of the corresponding parameters.

18. The computer-readable medium of claim 15 , wherein the software further causes the second processor to:

identify, as statistical outliers, values of a parameter from a subset of the plurality of operational configurations that each differs by more than the tolerance value from a calculated statistical norm associated with the same parameter for the plurality of operational parameters;

correlate the statistical outliers to actual system failures; and

select, as the at least one same parameter, one or more parameters from the subset comprising correlations to the actual system failures higher than correlations of other remaining parameters from the subset.