Determining a recurrent problem of a computer resource using signatures

• US 20070083513A1
• Filed: 10/12/2005
• Published: 04/12/2007
• Est. Priority Date: 10/12/2005
• Status: Active Grant

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company), Valtrus Innovations Limited (f/k/a Dolya Holdco 9 Limited) (Key Patent Innovations Limited)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Cohen, Ira, Goldszmidt, Moises, Kelly, Terence, Symons, Julie, Zhang, Steve
• Granted Patent: US 7,502,971 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 11/008 Reliability or availability analysis

Abstract

A computer system includes a signature creation engine operable to determine signatures representing states of a computer resource from metrics for the computer resource. The computer system also includes a database operable to store the signatures along with an annotation for each signature including information relating to a state of the computer resource. The computer system is operable to determine a recurrent problem of the computer resource from stored signatures.

Claims

1. A method of determining a recurrent problem of at least one computer resource, the method comprising:

determining signatures for at least one computer resource, each signature representing a state of the at least one computer resource;

automatically determining similarities between the signatures; and

determining whether a recurrent problem of the at least one computer resource exists based on at least some of the similarities.

2. The method of claim 1, wherein automatically determining similarities between the signatures further comprises:

determining groups of similar signatures.

3. The method of claim 2, wherein determining whether a recurrent problem of the at least one computer resource exists based on the similarities further comprises:

determining a time of occurrence of the signatures in each group; and

identifying signatures in the same group having metrics captured during service level objective violations for the at least one computer resource.

4. The method of claim 3, wherein the groups are clusters determined using a clustering algorithm and the method further comprises:

evaluating a purity of a cluster having the identified signatures and a similarity between signatures in the cluster.

5. The method of claim 4, further comprising:

identifying labels for the signatures in the cluster; and

determining the purity of the cluster based on similarities between the labels of the signatures in the cluster.

6. The method of claim 3, further comprising:

determining whether the recurrent problem comprises at least two different problems or a single recurrent problem based on whether signatures having metrics captured during the service level objective violations are included in a same group or in different groups.

7. The method of claim 2, wherein determining groups of similar signatures further comprises:

determining clusters of signatures;

determining a purity of each cluster;

determining whether the purity of any of the clusters is less than a predetermined purity; and

for each cluster determined to have a purity less than the predetermined purity, determining new clusters for signatures from the cluster.

8. The method of claim 7, further comprising:

determining a label for each of the signatures, wherein the label is representative of a state of a computer resource; and

identifying signatures having a same label in each cluster to determine the purity of each cluster.

9. The method of claim 7, wherein determining the purity based on the similarity further comprises:

determining an entropy for each cluster, wherein the entropy is based on probability estimates that signatures in a cluster refer to the same label.

10. The method of claim 9, wherein the label for each signature represents a violation of a service level objective or represents a compliance with the service level objective.

11. The method of claim 9, wherein the label for a signature in a cluster of the clusters is based on a centroid of the cluster.

12. The method of claim 11, wherein the centroid is associated with a mean of the cluster or a median of the cluster.

13. The method of claim 9, further comprising:

providing the label for each signature in an annotation for each signature.

14. The method of claim 7 wherein new clusters are determined until purity is at least as pure as the predetermined purity for all the clusters or until a total number of clusters exceeds a threshold.

15. The method of claim 7, wherein the purity for each cluster is a metric indicating the homogeneity of labels for signatures in a respective cluster and the method further comprises:

evaluating whether a cluster is representative of a single state of a computer resource based on the purity.

16. The method of claim 7, wherein determining clusters of signatures comprises:

using a distance metric to determine similarities between the signatures and to assign the signatures to clusters.

17. The method of claim 1, further comprising:

prioritizing problems of the at least one computer resource based on at least one of a frequency of occurrence of the problems and whether a problem has occurred in one computer resource or multiple computer resources.

18. The method of claim 1, further comprising:

storing annotations with the signatures, the annotations including descriptions of a service level objective violation and a solution for fixing the violation.

19. The method of claim 1, wherein the at least one computer resource comprises different computer systems operable to be located in different geographic areas.

20. The method of claim 1, wherein the recurrent problem comprises a performance problem.

21. The method of claim 1, wherein determining the signatures further comprises:

determining the signatures using at least one of raw values, metric attribution, and a combination of raw values and metric attribution.

22. The method of claim 1, further comprising:

collecting a plurality of metrics for the at least one computer resource; and

determining the signatures using metric attribution of the plurality of metrics, wherein the metric attribution includes determining whether a metric of the plurality of metrics is attributed to a service level objective violation.

23. A method of determining a syndrome of a computer resource problem, the method comprising:

grouping signatures of at least one computer resource based on similarities, wherein the signatures represent states of the at least one computer resource; and

for at least one of the groups, determining a common denominator that characterizes the signatures in the group, wherein the common denominator is representative of a syndrome of a problem of the at least one computer resource.

24. The method of claim 23, wherein grouping signatures of at least one computer resource based on similarities further comprises:

determining clusters of the signatures based on similarities of the signatures.

25. The method of claim 23, wherein determining a common denominator further comprises:

determining a centroid for each cluster including signatures associated with a problem of the at least one computer resource.

26. The method of claim 23, wherein determining a common denominator further comprises:

determining a most frequently occurring substring in the signatures in each cluster including signatures associated with a problem of the at least one computer resource.

27. The method of claim 23, wherein determining a common denominator further comprises:

determining a largest, most frequently occurring substring in the signatures in each group.

28. The method of claim 23, further comprising:

annotating signatures in a group with the syndrome of the problem of the at least one computer resource.

29. The method of claim 28, wherein annotating signatures further comprises:

generating a label including the syndrome of the problem of the at least one computer resource; and

annotating the signatures in the group with the label.

30. The method of claim 28, further comprising:

using at least one signature annotated with the syndrome of the problem of the at least one computer resource to diagnose a problem with another computer resource having a signature similar to the at least one annotated signature.

31. A computer system comprising:

a signature creation engine operable to determine signatures representing states of at least one computer resource from metrics for the at least one computer resource;

a recurrent problem identifier module operable to identify a recurrent problem of the at least one computer resource from the signatures; and

a database operable to store the signatures along with an annotation for each signature including information relating to a state of the at least one computer resource.

32. The computer system of claim 31, further comprising:

a common denominator module operable to determine a syndrome of a problem of the at least one computer resource from the signatures.

33. The computer system of claim 31, wherein the annotations include at least one of a diagnosis of a problem, a solution to the problem, configuration of a system, identification of a system administrator, a label identifying a problem associated with the signature, and a purity of a cluster including the signature.

34. The computer system of claim 31, further comprising:

an interface operable to allow a user to annotate the signatures with the annotations.

35. The computer system of claim 31, wherein the database is searchable to identify at least one signature closely matching a signature for another computer resource used to search the database and is operable to retrieve each closely matching signature and information annotated for each closely matching signature to diagnose a problem for the another computer resource.

36. A method comprising:

receiving a search signature including a signature for a first computer system having a problem; and

searching clusters of stored signatures for at least one signature similar to the search signature, wherein the signatures are clustered to group similar signatures and each signature includes an annotation including a description of a state of a computer system associated with the signature; and

retrieving the annotation for at least one of the similar signatures.

37. The method of claim 36, wherein searching clusters of stored signatures further comprises:

determining similarities between the search signature and at least one signature or a centroid of each cluster.

38. The method of claim 36, further comprising:

at least one of diagnosing the problem of the first computer and determining a solution to the problem using the retrieved annotation.

39. The method of claim 36, wherein the stored signatures comprise signatures associated with at least one other computer system different from the first computer system.

40. An apparatus comprising:

means for determining signatures representing states of at least one computer resource;

means for storing the signatures and an annotation for each signature including information associated with a state of the at least one computer resource.

means for determining a recurrent problem of the at least one computer resource from at least some of the signatures;

means for determining a syndrome of a problem of the at least one computer resource from at least some of the signatures; and

means for searching the stored signatures to identify at least one closely matching signature of a signature of another computer resource to at least one of diagnose and determine a solution of a problem of the another computer resource.

41. A computer program stored on a computer readable medium and comprising code that when executed causes a computer to perform the following:

determining signatures representing states of at least one computer resource;

identifying a recurrent problem of the at least one computer resource from at least some of the signatures; and

determining a syndrome of the recurrent problem from a group of the signatures including the at least some of the signatures.