Drive failure prediction techniques for disk drives

• US 5,828,583 A
• Filed: 04/18/1996
• Issued: 10/27/1998
• Est. Priority Date: 08/21/1992
• Status: Expired due to Term

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Compaq Computer Corporation (HP Inc.)
• Inventors: Bush, Kenneth L., Lenny, Thomas R., Didner, Jonathan R.
• Primary Examiner(s): Trammell, James P.
• Application Number:
  US08/635,429
• Time in Patent Office: 922 Days
• Field of Search: 360/75, 360/103, 360/104, 360/137, 364/551.01, 395/575, 395/183.01, 395/183.02, 395/183.18, 395/184.01, 371/3, 371/5.1-5.5, 371/21.2, 371/21.4, 371/21.6, 371/25.1, 371/26
• US Class Current: 702/185
• CPC Class Codes: G06F 11/004 Error avoidance G06F11/07 and subgroups take precedence
• CPC Class Codes: G06F 11/008 Reliability or availability analysis
• CPC Class Codes: H04L 41/00 Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
• CPC Class Codes: H04L 41/0213 Standardised network management protocols, e.g. simple network management protocol [SNMP]
• CPC Class Codes: H04L 41/0233 Object-oriented techniques, for representation of network management data, e.g. common object request broker architecture [CORBA]
• CPC Class Codes: H04L 9/40 Network security protocols

Abstract

Method for predicting an imminent failure of a disk drive. A plurality of attributes are selected for monitoring during operation of the disk drive. These attributes may include self-preserving attributes, performance attributes, error rate attributes, and even count attributes. An initial value is determined for each one of the selected attributes. A threshold value is then selected for each of the attributes. The disk drive is then monitored for occurrences of the self-preserving, performance, error rate and event count attributes. Each time an occurrence of the attributes is monitored, the initial value for that attribute is updated and normalized. The updated normalized value for the attribute is then compared with the corresponding threshold for that attribute and an imminent failure of the disk drive is predicted if the normalized updated value of the attribute exceeds the threshold therefor.

Claims

1. A method of predicting an imminent failure of a storage device within a computer system, comprising:

selecting an attribute as indicative of an imminent failure of a storage device, said selected attribute being a performance attribute;

monitoring said selected performance attribute during normal operation of said storage device;

selecting threshold performance speed for said selected performance attribute;

automatically predicting an imminent failure of said storage device based upon determining that said selected performance attribute has a speed less than said threshold performance speed; and

alerting an operator of the computer system of the imminent failure of said storage device.

2. A method of predicting an imminent failure of a storage device within a computer system, comprising:

selecting an attribute as indicative of an imminent failure of a storage device;

monitoring said selected attribute during normal operation of said storage device;

automatically predicting an imminent failure of said storage device based upon changes in said monitored attribute; and

alerting an operator of the computer system of the imminent failure of said storage device,

wherein said selected attribute is an error rate attribute and wherein said predicting an imminent failure of said storage device based upon changes in said selected attribute operates to predict an imminent failure of said storage device upon detecting an increase in said error rate attribute.

3. A method of predicting an imminent failure of a storage device according to claim 2 and further comprising:

selecting a threshold error rate for said selected error rate attribute; and

predicting an imminent failure of said storage device upon determining that said monitored error rate attribute has an error rate greater than said threshold error rate.

4. A method of predicting an imminent failure of a storage device within a computer system, comprising:

selecting an attribute as indicative of an imminent failure of a storage device;

monitoring said selected attribute during normal operation of said storage device;

normalizing said monitored attribute;

automatically predicting an imminent failure of said storage device based upon changes in said normalized monitored attribute; and

alerting an operator of the computer system of the imminent failure of said storage device,

wherein said selected attribute is an event count attribute and wherein said predicting an imminent failure of said storage device based upon changes in said normalized monitored attribute operates to predict an imminent failure of said storage device upon detecting an increase in said event count attribute.

5. A method of predicting an imminent failure of a storage device according to claim 4 and further comprising:

selecting a threshold event count for said selected event count attribute; and

predicting an imminent failure of said storage device upon determining that said normalized monitored event count attribute has an event count greater than said threshold event count.

6. For a computer system having a host computer and a storage device coupled to said host computer, a method of predicting an imminent failure of said storage device, comprising:

selecting at least one attribute for monitoring during operation of said storage device;

selecting a corresponding threshold value for each one of said at least one attribute, said selecting of the corresponding threshold value includes at least,

retrieving, from said storage device, a threshold value for each one of said at least one attribute, and

storing, in said volatile memory of said host computer, said retrieved threshold value for each one of said at least one attribute;

obtaining a previously stored initial value for each one of said at least one attribute;

monitoring said storage device to detect occurrences of said at least one attribute;

upon detection of an occurrence of one of said at least one attribute, updating said initial value for said attribute for which said occurrence has been detected; and

automatically predicting an imminent failure of said storage device if said updated value of said attribute for which said occurrence has been detected crosses said corresponding threshold value for said attribute.

7. A method of predicting an imminent failure of a storage device according to claim 6, wherein said host computer further comprises a volatile memory and said storage device comprises a drive attribute values sector, and

wherein said obtaining of said previously stored initial value for each one of said at least one attribute comprises:

retrieving, from said drive attribute values sector of said storage device, a saved attribute value for each one of said at least one attribute; and

storing, in said volatile memory of said host computer, said retrieved saved attribute value for each one of said at least one attribute as a current value for each one of said at least one attribute.

8. A method of predicting an imminent failure of a storage device according to claim 7 wherein said storage device further comprises a warranty failure thresholds sectors, and

wherein said retrieving of said threshold value for each one of said at least one attribute operates to retrieve said threshold values from said warranty failure thresholds sector of said storage device.

9. A method of predicting an imminent failure of a storage device according to claim 7 and further comprising:

monitoring operations of said storage device to detect an occurrence of a first one of said at least one attribute; and

upon detecting said occurrence of said first one of said at least one attribute, updating, in said volatile memory, a first current value which corresponds to said first one of said at least one attribute.

10. A method of predicting an imminent failure of a storage device according to claim 9 and further comprising:

periodically overwriting said saved attribute value for each one of said at least one attribute stored in said drive attribute values sector of said storage device with said current value of each one of said at least one attribute stored in said volatile memory.

11. A method of predicting an imminent failure of a storage drive within a computer system, comprising:

selecting an attribute as indicative of an imminent failure of a storage device;

monitoring said selected attribute during normal operation of said storage device;

automatically predicting an imminent failure of said storage device based upon changes in said selected attribute, said predicting including at least,

selecting a threshold value for the selected attribute, and

predicting an imminent failure of said storage device upon determining that said selected attribute crosses said selected threshold value, said predicting including at least (i) normalizing said selected attribute for which said occurrence was monitored, and (ii) comparing said normalized selected attribute with the threshold value corresponding thereto; and

alerting an operator of the computer system of the imminent failure of said storage device.

12. A method of predicting an imminent failure of a storage drive within a computer system, comprising:

selecting an attribute as indicative of an imminent failure of a storage device;

monitoring said selected attribute during normal operation of said storage device;

automatically predicting an imminent failure of said storage device based upon changes in said selected attribute, said predicting including at least,

selecting a threshold value for the selected attribute, the selected threshold value being a warranty failure threshold, and

predicting an imminent failure of said storage device upon determining that said selected attribute crosses said selected threshold value; and

alerting an operator of the computer system of the imminent failure of said storage device.

13. For a computer system having a host computer and a storage device coupled to said host computer, a method of predicting an imminent failure of a storage device, comprising:

selecting at least one attribute for monitoring during operation of said storage device;

selecting a corresponding threshold value for each one of said at least one attribute;

obtaining a previously stored initial value for each one of said at least one attribute;

monitoring said storage device to detect occurrences of said at least one attribute;

upon detection of an occurrence of one of said at least one attribute, updating said initial value for said attribute for which said occurrence has been detected; and

automatically predicting an imminent failure of said storage device if said updated value of said attribute for which said occurrence has been detected crosses said corresponding threshold value for said attribute, said predicting including at least,

normalizing said updated value for said attribute for which said occurrence was monitored, and

comparing said normalized updated value with said corresponding threshold for said attribute.

14. A method of predicting an imminent failure of a storage device according to claim 13, wherein said method further comprises:

alerting an operator of the computer system of the imminent failure of said storage device.

15. For a computer system having a host computer and a storage device coupled to said host computer, a method of predicting an imminent failure of a storage drive, comprising:

selecting at least one attribute for monitoring during operation of said storage device;

selecting a corresponding threshold value for each one of said at least one attribute, said corresponding threshold value being a warranty failure threshold;

obtaining a previously stored initial value for each one of said at least one attribute;

monitoring said storage device to detect occurrences of said at least one attribute;

upon detection of an occurrence of one of said at least one attribute, updating said initial value for said attribute for which said occurrence has been detected; and

automatically predicting an imminent failure of said storage device if said updated value of said attribute for which said occurrence has been detected crosses said corresponding threshold value for said attribute.

16. A method of predicting an imminent failure of a storage device according to claim 15, wherein said method further comprises:

alerting an operator of the computer system of the imminent failure of said storage device.

17. A method of predicting an imminent failure of a disk drive, comprising the steps of:

selecting a plurality of attributes for monitoring during operation of said disk drive, said plurality of attributes including at least one self-preserving attribute, at least one performance attribute, at least one error rate attribute and at least one event count attribute;

determining an initial value for each one of said at least one self-preserving attribute, at least one performance attribute, at least one error rate attribute and at least one event count attribute;

selecting a threshold value for each of said at least one self-preserving attribute, a threshold performance speed for each of said at least one performance attribute, a threshold error rate for each of said at least one error rate attribute and a threshold event count for each of said at least one event count attribute;

if disk drive failure prediction is enabled, monitoring said disk drive for occurrences of said at least one self-preserving attribute, at least one performance attribute, at least one error rate attribute and at least one event count attribute;

if disk drive failure prediction is not enabled, monitoring said disk drive for occurrences of said at least one self-preserving attribute;

each time an occurrence of said at least one self-preserving attribute, said at least one performance attribute, said at least one error rate attribute or said at least one event count attribute is monitored, updating said initial value for said attribute for which said occurrence was monitored; and

predicting an imminent failure of said disk drive if said updated value of said attribute for which said occurrence was monitored exceeds a corresponding threshold for said attribute.

18. A method of predicting an imminent failure of a disk drive according to claim 17 wherein the step of predicting an imminent failure of said disk drive if said updated value of said attribute for which said occurrence was monitored exceeds a corresponding threshold for said attribute further comprises the steps of:

normalizing said updated value for said attribute for which said occurrence was monitored; and

comparing said normalized updated value with said corresponding threshold for said attribute.

19. A method of predicting an imminent failure of a disk drive according to claim 17 wherein said attribute for which said occurrence was monitored was one of said at least one event count attribute and wherein the step of predicting an imminent failure of said disk drive if said updated value of said attribute for which said occurrence was monitored exceeds a corresponding threshold for said attribute further comprises the steps of:

comparing said updated value of said event count attribute for which said occurrence was monitored to said corresponding threshold event count; and

predicting an imminent failure of said disk drive if said updated value of said event count attribute for which said occurrence was monitored exceeds said corresponding threshold event count.

20. A method of predicting an imminent failure of a disk drive according to claim 19 wherein said at least one event count attribute includes a re-allocated sector count attribute, a raw read error count attribute, spin retry count attribute, a read channel margin attribute, a drive calibration retry count and a drive power cycle count attribute.

21. A method of predicting an imminent failure of a disk drive according to claim 17 wherein said attribute for which said occurrence was monitored was one of said at least one performance attribute and wherein the step of predicting an imminent failure of said disk drive if said updated value of said attribute for which said occurrence was monitored exceeds a corresponding threshold for said attribute further comprises the steps of:

comparing said updated value of said performance attribute for which said occurrence was monitored to said corresponding threshold performance speed; and

predicting an imminent failure of said disk drive if said updated value of said performance attribute for which said occurrence was monitored exceeds said corresponding threshold performance speed.

22. A method of predicting an imminent failure of a disk drive according to claim 21 wherein said at least one performance attribute includes a spin up time attribute, a throughput performance attribute and a seek time performance attribute.

23. A method of predicting an imminent failure of a disk drive according to claim 17 wherein said attribute for which said occurrence was monitored was one of said at least one error rate attribute and wherein the step of predicting an imminent failure of said disk drive if said updated value of said attribute for which said occurrence was monitored exceeds a corresponding threshold for said attribute further comprises the steps of:

comparing said updated value of said error rate attribute for which said occurrence was monitored to said corresponding threshold error rate; and

predicting an imminent failure of said disk drive if said updated value of said error rate attribute for which said occurrence was monitored exceeds said corresponding threshold error rate.

24. A method of predicting an imminent failure of a disk drive according to claim 23 wherein said at least one error rate attribute includes a seek error rate attribute.

25. A method of predicting an imminent failure of a disk drive according to claim 17 wherein said attribute for which said occurrence was monitored was one of said at least one self-preserving attribute and wherein the step of predicting an imminent failure of said disk drive if said updated value of said attribute for which said occurrence was monitored exceeds a corresponding threshold for said attribute further comprises the steps of:

comparing said updated value of said self-preserving attribute for which said occurrence was monitored to said corresponding threshold value; and

predicting an imminent failure of said disk drive if said updated value of said self-preserving attribute for which said occurrence was monitored exceeds said corresponding threshold value.

26. A method of predicting an imminent failure of a disk drive according to claim 25 wherein said at least one self-preserving attribute includes a power-on hours count attribute and a start/stop count attribute.

27. A method of predicting an imminent failure of a disk drive according to claim 17 and further comprising the step of performing off-line analysis of said disk drive.

28. A method of predicting an imminent failure of a storage device according to claim 27 wherein said monitoring of said disk drive for occurrences of said at least one self-preserving attribute, at least one error rate attribute and at least one event count attribute continues during said off-line analysis of said disk drive.

29. A method of predicting an imminent failure of a disk drive according to claim 28 wherein said disk drive further comprises command block registers which include a cylinder low register and a cylinder high register and wherein the step of performing off-line analysis of said disk drive further comprises the steps of:

placing a key code in said cylinder low register and said cylinder high register; and

performing off-line analysis of said disk drive only if said key code has been placed in said cylinder low and said cylinder high register;

wherein said key code placed in said cylinder low and cylinder high registers do not match any cylinder address for said disk drive.

30. A method of predicting an imminent failure of a storage device according to claim 17 wherein said disk drive is an ATA disk drive.

31. A method of executing a command for a disk drive having interface I/O registers which include a command register, a cylinder low register and a cylinder high register, comprising the steps of:

placing a command in said command register;

placing a first key code in said cylinder low register and a second key code in said cylinder high register; and

executing said command only if said first key code has been placed in said cylinder low register and said second key code has been placed in said cylinder high register;

wherein said first key code placed in said cylinder low and said second key code placed in said cylinder high registers do not match any cylinder address for said disk drive.

32. A method of executing a command for a disk drive according to claim 31 wherein said command is an execute drive failure prediction command.

33. A method of predicting an imminent failure of a storage device according to claim 31 wherein said disk drive is an ATA disk drive and said interface I/O registers are ATA interface I/O registers.