Automatic software fault diagnosis by exploiting application signatures

1. A method of automating runtime failure analysis for a computer application: operating within a runtime environment, the method comprising:

monitoring non-faulty application traces of said computer application with said runtime environment during fault-free activities to create non-faulty runtime signatures for said computer application;

storing said non-faulty runtime signatures;

detecting a faulty application trace of said computer application,

comparing said faulty application trace with said non-faulty runtime signatures to identify fault attributes; and

outputting said fault attributes.

2. The method according to claim 1 , said monitoring of said non-faulty application traces provides non-faulty attributes of said computer application, said method further comprising:

sorting said non-faulty attributes into groups of attributes;

merging said groups of attributes to form said application signatures; and

removing runtime variants from said application signatures.

3. The method according to claim 1 , said monitoring of said non-faulty application traces provides information regarding system calls, including system call number, parameters, return value and error number; signal numbers; environment names and values; resource limits; access control attributes including user and group identification values; and a call graph of system calls.

4. The method according to claim 1 , said monitoring of said non-faulty application traces comprises generating a call graph of system calls in a process comprising:

adding a node to said call graph for each system call based on a system call memory context indicating from where within said computer application each system call was invoked;

adding an arc from a parent node to each newly added node; and

merging nodes of the same system call type having the same parent node.

5. The method according to claim 1 , said comparing comprising comparing a call graph of said faulty application trace with call graphs of non-faulty application traces of said application signatures.

6. The method according to claim 1 , further comprising ranking said fault attributes found by assigning a higher weight to fault attributes found earlier relative to fault attributes found later.

7. A method of automating runtime failure analysis for a computer application operating within a runtime environment, the method comprising:

monitoring non-faulty application traces of said computer application with said runtime environment during fault-free activities to create non-faulty runtime signatures for said computer application, said non-faulty application traces comprising interactions between said computer application and said runtime environment, said non-faulty application traces being gathered during multiple runs of said computer application;

storing said non-faulty runtime signatures;

detecting a faulty application trace of said computer application, said faulty application trace comprising interactions between said computer application and said runtime environment;

comparing said faulty application trace with said non-faulty runtime signatures to identify fault attributes; and

outputting said fault attributes.

8. The method according to claim 7 , said monitoring of said non-faulty application traces provides non-faulty attributes of said computer application, said method further comprising:

sorting said non-faulty attributes into groups of attributes;

merging said groups of attributes to form said application signatures; and

removing runtime variants from said application signatures.

9. The method according to claim 7 , said monitoring of said non-faulty application traces provides information regarding system calls, including system call number, parameters, return value and error number; signal numbers; environment names and values; resource limits; access control attributes including user and group identification values; and a call graph of system calls.

10. The method according to claim 7 , said monitoring of said non-faulty application traces comprises generating a call graph of system calls in a process comprising:

adding a node to said call graph for each system call based on a system call memory context indicating from where within said computer application each system call was invoked;

adding an arc from a parent node to each newly added node; and

merging nodes of the same system call type having the same parent node.

11. The method according to claim 7 , said comparing comprising comparing a call graph of said faulty application trace with call graphs of non-faulty application traces of said application signatures.

12. The method according to claim 7 , further comprising ranking said fault attributes found by assigning a higher weight to fault attributes found earlier relative to fault attributes found later.

13. A method of automating runtime failure analysis for a computer application operating within a runtime environment, the method comprising:

monitoring non-faulty application traces of said computer application with said runtime environment during fault-free activities to create non-faulty runtime signatures for said computer application, said non-faulty application traces comprising interactions between said computer application and said runtime environment, said non-faulty application traces being gathered during multiple runs of said computer application;

storing said non-faulty runtime signatures;

detecting a faulty application trace of said computer application, said faulty application trace comprising interactions between said computer application and said runtime environment;

comparing said faulty application trace with said non-faulty runtime signatures by comparing a call graph of said faulty application trace with call graphs of non-faulty application traces of said application signatures to identify fault attributes;

ranking said fault attributes based on: giving more weight to fault attributes that are invariants and comprise fault attributes with a constant value; and giving more weight to fault attributes located closer to a head of said call graph of said faulty application trace; and

outputting said fault attributes.

14. The method according to claim 13 , said monitoring of said non-faulty application traces provides non-faulty attributes of said computer application, said method further comprising:

sorting said non-faulty attributes into groups of attributes;

merging said groups of attributes to form said application signatures; and

removing runtime variants from said application signatures.

15. The method according to claim 13 , said monitoring of said non-faulty application traces provides information regarding system calls, including system call number, parameters, return value and error number; signal numbers; environment names and values; resource limits; access control attributes including user and group identification values; and a call graph of system calls.

16. The method according to claim 13 , said monitoring of said non-faulty application traces comprises generating a call graph of system calls in a process comprising:

adding a node to said call graph for each system call based on a system call memory context indicating from where within said computer application each system call was invoked;

adding an arc from a parent node to each newly added node; and

merging nodes of the same system call type having the same parent node.

17. A computer program storage device comprising a computer storage medium readable by computer, storing instructions executable by said computer to perform a method of automating runtime failure analysis for a computer application operating within a runtime environment, said method comprising:

monitoring non-faulty application traces of said computer application with said runtime environment during fault-free activities to create non-faulty runtime signatures for said computer application;

storing said non-faulty runtime signatures;

detecting a faulty application trace of said computer application,

comparing said faulty application trace with said non-faulty runtime signatures to identify fault attributes; and

outputting said fault attributes.

18. The computer program storage device according to claim 17 , said monitoring of said non-faulty application traces provides non-faulty attributes of said computer application, said method further comprising:

sorting said non-faulty attributes into groups of attributes;

merging said groups of attributes to form said application signatures; and

removing runtime variants from said application signatures.

19. The computer program storage device according to claim 17 , said monitoring of said non-faulty application traces provides information regarding system calls, including system call number, parameters, return value and error number; signal numbers; environment names and values; resource limits; access control attributes including user and group identification values; and a call graph of system calls.

20. The computer program storage device according to claim 17 , said monitoring of said non-faulty application traces comprises generating a call graph of system calls in a process comprising:

adding a node to said call graph for each system call based on a system call memory context indicating from where within said computer application each system call was invoked;

adding an arc from a parent node to each newly added node; and

merging nodes of the same system call type having the same parent node.