Systems and methods for automated management of software application deployments

• US 7,870,550 B1
• Filed: 12/21/2005
• Issued: 01/11/2011
• Est. Priority Date: 12/21/2004
• Status: Active Grant

• Current Assignee: Citrix Systems, Inc. (Cloud Software Group)
• Original Assignee: Zenprise, Inc. (Cloud Software Group)
• Inventors: Hassan, Tanvir, Roach, Kelly Brian, Qureshi, Waheed
• Primary Examiner(s): Zhen; Wei Y
• Assistant Examiner(s): Brophy; Matthew J
• Application Number:
  US11/316,202
• Time in Patent Office: 1,847 Days
• Field of Search: 717124-135, 717168-178
• US Class Current: 717/174
• CPC Class Codes: G06F 11/079 Root cause analysis, i.e. error or fault diagnosis in a hardware test environment G06F11/22; in a software test environment G06F11/36
• CPC Class Codes: G06N 5/048 Fuzzy inferencing

Abstract

Systems and methods for managing a deployment of a software application are disclosed. One system comprises a machine-readable repository of rules describing problems associated with the software application, a monitoring component that collects state metrics from the deployment, and a problem detector that programmatically uses the rules to map combinations of the collected state metrics to problems.

Claims

1. A meta-application system for managing a deployment of a software application, comprising:

a computer system comprising one or more computers, said computer system programmed to implement at least;

a non-transitory machine-readable repository of encoded knowledge comprising;

a collection of deployment-independent rules describing problems associated with the software application, at least some of the rules comprising logical combinations of conditions individually satisfied by features of the deployment; and

deployment-independent remedies for countering the problems;

wherein the rules collectively comprise predicates, logical operators, mathematical operators, free variables, and constant terms, the logical operators including AND and OR operators;

a monitoring component that collects data values of state metrics from the deployment, wherein the meta-application system is configured to derive at least one of the features from one or more of the collected state metric data values;

and a problem detector that detects instances of the problems within the deployment, at least by programmatically using the features of the deployment to satisfy the logical combinations of conditions of the rules, the problem detector configured to satisfy at least some of the conditions by conducting pattern-matching between features represented as closed atomic formulae with constant term arguments and no free variables, and conditions represented as open atomic formulae with free variables.

2. The meta-application system of claim 1, further comprising a feature detector that analyzes collected streams of the state metric data values to identify at least some of the features in the behavior of the deployment.

3. The meta-application system of claim 1, further comprising a remedy planning module configured to convert the remedies into deployment-specific plans.

4. The meta-application system of claim 3, further comprising a remedy execution engine configured to automatically execute the plans.

5. The meta-application system of claim 1, further comprising an application model comprising:

representation of the configuration and behavior of the deployment; and

a query interface allowing other components of the system to query the application model for information contained within the application model.

6. The meta-application system of claim 5, wherein the system dynamically updates the application model.

7. The meta-application system of claim 5, wherein the problem detector is configured to use information from the application model as one or more of the features to evaluate whether the deployment has one of the problems identified by the encoded knowledge.

8. The meta-application system of claim 5, wherein the representation of the configuration and behavior of the deployment comprises an object graph, and wherein objects of the object graph represent at least hardware and software elements of the deployment, the meta-application system further comprising a root cause analysis module configured to correlate problems partially or completely detected by the problem detector with objects of the object graph, the root cause analysis module also configured to identify possible root causes of the partially or completely detected problems by conducting pattern-recognition against portions of the object graph that include at least some of the correlated objects.

9. The meta-application system of claim 5, further comprising a remedy planning module configured to use information from the application model to convert the remedies into deployment-specific plans.

10. The meta-application system of claim 9, further comprising a remedy execution engine that automatically executes the plans.

23. The meta-application system of claim 8, wherein the root cause analysis module is configured to report causes that the root cause analysis module has ruled out as possible root causes of the completely or partially detected problems.

24. The meta-application system of claim 10, wherein the remedy execution engine is configured to compute an optimal sequence in which to execute a plurality of the plans.

26. The meta-application system of claim 1, wherein the repository of encoded knowledge is configured to receive and store updates sent through a computer network from one or more update servers, the updates comprising additions, modifications, and/or deletions of the encoded knowledge.

27. The meta-application system of claim 26, wherein the knowledge updates comprise new deployment-independent rules describing problems associated with the software application, at least some of the new rules comprising logical combinations of conditions individually satisfied by features of the deployment.

28. The meta-application system of claim 1, wherein at least one of the rules is configured to be used by the problem detector to construct a circuit comprising:

a plurality of atomic gates each configured to be used by the problem detector to test whether one of said features satisfies one of the conditions of the rule, and to produce an output bead if the condition is satisfied; and

a plurality of operator gates each having one or more inputs, each of the inputs configured to receive output beads from one of the atomic gates or from another one of the operator gates, each operator gate configured to be used by the problem detector to test for a condition associated with received beads and to produce an output bead if the condition is met;

wherein the rule is configured to be applied by the problem detector to automatically detect one of the problems within the deployment based on output beads generated from use of the operator gates.

29. The meta-application system of claim 1, wherein the meta-application system is configured to customize the encoded knowledge based on feedback data collected over time from the deployment.

30. The meta-application system of claim 1, wherein at least one of the conditions of the collection of rules is numerical, at least some of the state metric data values collected by the monitoring component comprising numerical values, the meta-application system being configured to derive at least one of the features from one or more of the collected numerical state metric data values.

11. A computer-implemented method of managing a deployment of a software application, comprising:

collecting data values of state metrics from the deployment;

accessing a machine-readable repository of encoded knowledge, the repository comprising;

a collection of deployment-independent rules describing problems associated with the software application, at least some of the rules comprising logical combinations of conditions individually satisfied by features of the deployment;

and deployment-independent remedies for countering the problems;

wherein the rules collectively comprise predicates, logical operators, mathematical operators, free variables, and constant terms, the logical operators including AND and OR operators;

deriving at least one of the features from one or more of the collected state metric data values; and

detecting instances of the problems within the deployment, at least by programmatically using the features of the deployment to satisfy at least some of the logical combinations of conditions of the rules, wherein satisfying at least some of the conditions comprises conducting pattern-matching between features represented as closed atomic formulae with constant term arguments and no free variables, and conditions represented as open atomic formulae with free variables;

wherein said programmatically using the features includes using the derived feature, the method being performed by one or more computing devices.

12. The method of claim 11, further comprising analyzing collected streams of the state metric data values to identify at least some of the features in the behavior of the deployment.

13. The method of claim 11, further comprising converting a remedy associated with a detected problem into a deployment-specific plan.

14. The method of claim 13, further comprising automatically executing the plan.

15. The method of claim 11, further comprising providing an application model comprising:

a representation of the deployment; and

a query interface allowing other components to query the application model for information contained within the application model.

16. The method of claim 15, further comprising dynamically updating the application model.

17. The method of claim 15, further comprising using information from the application model as one or more of the features to evaluate whether the deployment has one of the problems identified by the encoded knowledge.

18. The method of claim 15, wherein the representation of the deployment comprises an object graph, and wherein objects of the object graph represent at least hardware and software elements of the deployment, the method further comprising:

correlating problems detected by the problem detector with objects of the object graph;

and identifying possible root causes of the detected problems by conducting pattern-recognition against portions of the object graph that include at least some of the correlated objects.

19. The method of claim 15, further comprising using information from the application model to convert the remedies into deployment-specific plans.

20. The method of claim 19, further comprising automatically executing one or more of the plans.

21. The method of claim 20, further comprising computing an optimal sequence in which to execute a plurality of the plans.

22. The method of claim 18, further comprising reporting causes that have been ruled out as possible root causes of the detected problems.

25. A meta-application system for managing a deployment of a software application, comprising:

a computer system comprising one or more computers, said computer system programmed to implement at least;

a monitoring component that collects data values of state metrics from the deployment;

a problem detector that programmatically uses features of the deployment to at least partially satisfy predefined logical combinations of conditions, said logical combinations being deployment-independent, the problem detector configured to satisfy at least some of the conditions by conducting pattern-matching between features represented as closed atomic formulae with constant term arguments and no free variables, and conditions represented as open atomic formulae with free variables;

a feature detector that uses the collected state metric data values to detect at least some of the features;

an application model comprising;

an object graph representation of the configuration and behavior of the deployment, wherein objects of the object graph represent at least hardware and software elements of the deployment; and

a query interface allowing other components of the meta-application system to query the application model for information contained within the application model;

and a root cause analysis module configured to correlate problems partially or completely detected by the problem detector with objects of the object graph, the root cause analysis module also configured to identify possible root causes of the partially or completely detected problems by conducting pattern-recognition against portions of the object graph that include at least some of the correlated objects.

31. The meta-application system of claim 25, wherein said pattern-recognition comprises:

comparing a candidate object graph object that is one of the correlated objects with other object graph objects that are of an identical type to that of the candidate object, to find a component of the candidate object that is dissimilar to corresponding components of the other objects with which the candidate object is compared; and

designating a deployment element corresponding to the dissimilar component as a root cause candidate.

32. The meta-application system of claim 25, wherein said pattern-recognition comprises:

finding a candidate object graph object that is one of the correlated objects and that has had a recent configurational change compared to nearby objects in the object graph; and

designating a deployment element corresponding to the candidate object as a root cause candidate.

33. The meta-application system of claim 25, wherein said pattern-recognition comprises grouping the correlated objects into clusters by k-means clustering, wherein numerical measurements from the deployment are used as cluster analysis dimensions.

34. The meta-application system of claim 25, wherein said pattern-recognition comprises:

finding a candidate object graph object that is one of the correlated objects and that has directed links to a plurality of the correlated objects;

determining that none of the correlated objects have directed links to the candidate object; and

designating a deployment element corresponding to the candidate object as a root cause candidate.

35. The meta-application system of claim 25, wherein said pattern-recognition comprises:

finding a candidate object graph object that has a directed link to one or more of the correlated objects;

determining that none of the correlated objects have directed links to the one or more correlated objects to which the candidate object has directed links; and

designating a deployment element corresponding to the candidate object as a root cause candidate.

36. The meta-application system of claim 25, wherein said pattern-recognition comprises:

finding a first candidate object graph object having a directed link to one or more of the correlated objects;

finding a second candidate object graph object having directed links to the one or more correlated objects and directed links to one or more non-correlated objects of the object graph; and

designating a deployment element corresponding to the first candidate object as a root cause candidate.

37. The meta-application system of claim 25, wherein said pattern-recognition comprises:

providing a temporal history of a state of one of the correlated objects of the object graph; and

using the temporal history to compare a current state of the correlated object with one or more previous states of the correlated object.