Identifying memory leaks in computer systems

• US 20060206885A1
• Filed: 03/10/2005
• Published: 09/14/2006
• Est. Priority Date: 03/10/2005
• Status: Active Grant

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Seidman, David Isaiah, Findeisen, Piotr
• Granted Patent: US 7,434,206 B2
• Time in Patent Office: Days
• US Class Current: 717/148
• CPC Class Codes: G06F 11/073 in a memory management context, e.g. virtual memory or cache management memory management G06F12/00; testing of static memory units G11C29/00
• CPC Class Codes: G06F 11/076 by exceeding a count or rate limit, e.g. word- or bit count limit
• CPC Class Codes: G06F 12/0253 Garbage collection, i.e. reclamation of unreferenced memory

Abstract

Embodiments of the invention relate to identifying sources of memory leaks. In an embodiment, an agent collects information related to object allocations and de-allocations in accordance with a sampling period, and, based on the relationship between object allocations and de-allocations, the agent analyzes the data and reports the results. Other embodiments are also disclosed.

Claims

1. A method for identifying a source of memory leak in a program, comprising:

while the program is running

obtaining information related to memory allocation of objects;

obtaining information related to memory de-allocation of the objects;

providing information related to the source of the memory leak based on one or a combination of

first difference in ratios of object allocation and de-allocation for a class;

second difference between a number of allocations and a number of de-allocations of an object;

a change in the first difference over time during execution of the program;

a change in the second difference over time during execution of the program;

a change in the first difference after a transaction of the program is executed;

a change in the second difference after a transaction of the program is executed;

a change in a heap size used by the program during its execution; and

similarity in the number of allocations of the object and a leak rate;

wherein the step of obtaining the information related to memory allocation is performed in one or a combination of

adding additional code to a class file that causes the information to be provided when the class file is being executed; and

requesting an interface to provide the information.

2. The method of claim 1 wherein adding the additional code occurs when the class file is about to be loaded into memory.

3. The method of claim 1 further comprising the step of tagging the object wherein, based on tagging, the information related to de-allocation of the object may be obtained.

4. The method of claim 1 wherein the information related to the allocation and/or the information related to the de-allocation is obtained based on a sampling period.

5. The method of claim 1 wherein the information related to the allocation and/or the information related to the de-allocation is based on a percentage of time spending in collecting the information and time spending executing the program.

6. The method of claim 1 wherein, upon initialization of a Java Virtual Machine, an agent registers with the Java Virtual Machine events that the agent is interested in being notified of.

7. The method of claim 1 wherein the program is written in the Java language, and the class file includes Java bytecode.

8. The method of claim 1 wherein the information related to the memory allocation and the information related to the memory de-allocation are obtained after the program reaches a steady state.

9. A system for providing information, comprising:

a Java Virtual Machine;

an agent including computer-executable instructions;

a processor processing data stored in memory; and

a class file;

wherein

the Java Virtual Machine provides the class file to the agent when the class file is about to be loaded into the memory;

the agent adds additional code into the class file in order to collect information related to memory allocation of an object based on a frequency including a first period of collecting the information and a second period of skipping collecting the information; and

the agent returns the instrumented class file to the Java Virtual Machine for it to load the instrumented class file into the memory;

while the instrumented class file is being executed by the processor the additional code causes the information to be provided; and

the information is for use in identifying a source of memory leak.

10. The system of claim 9 wherein an amount of the information is based on a percentage of time collecting the information and time executing a program using the class file.

11. The system of claim 10 wherein the percentage is based on a calculated error rate caused by the frequency.

12. The system of claim 9 wherein the agent is linked into the Java Virtual Machine during initialization of the Java Virtual Machine.

13. The system of claim 9 wherein the agent registers with the Java Virtual Machine events that the agent desires the Java Virtual Machine to notify the agent.

14. The system of claim 9 further comprising a garbage collector providing information related to object de-allocation.

15. A computer-readable medium embodying instructions for a computer to perform a method for identifying a source of a memory leak in a program, the method comprising:

based on a sampling frequency,

collecting first information related allocations of objects, and

collecting second information related to de-allocations of the objects; and

for a class

determining an allocation ratio for the class;

determining a de-allocation ratio for the class;

determining a difference between the allocation ratio and the de-allocation ratio;

comparing the difference against a threshold; and

based on results of comparing, determining whether objects of the class are part the source of the memory leak.

16. The computer-readable medium of claim 15 wherein the threshold is based on an error rate that is based on a number of sampling data.

17. The computer-readable medium of claim 16 wherein the error rate decreases when the program continues to be executed.

18. A method for determining sources of memory leaks in a program, comprising:

associating each data type to a confidence level that an object is a component of the memory leaks;

wherein a confidence level based on two or more data types is based on two or more confidence levels each associated with a data type of the two or more data types; and

while the program is running, based on a sampling frequency, collecting a plurality of data types related to allocations and de-allocations of objects of the program; and

based on the collected data types, providing confidence levels associated with the data types.

19. The method of claim 18 wherein the sampling frequency is based on one or a combination of

an error rate;

and

a percentage of time spending to collect the data types over time executing the program.

20. The method of claim 18 wherein a data type is selected from:

first difference in a ratio of object allocation and a ratio of object de-allocation of a class;

a change in the first difference over time while the program is running;

a change in the first difference after a transaction of the program is executed;

second difference in a number of allocations and a number of de-allocation for an object;

a change in the second difference over time while the program is running;

a change the second difference after a transaction of the program is executed;

a change in a heap size of the program;

similarity in a number of allocations of an object and a leak rate.

21. A system comprising:

means for determining a memory leak rate of a program;

means for determining a number of allocations of an object;

means for determining whether the object is a component of the leak rate;

wherein

determining whether the object is a component of the leak rate is based on the memory leak rate and the number of allocations of the object; and

all determining is performed while the program is being executed.

22. The system of claim 21 wherein determining whether the object is a component of the leak rate is based further on one or a combination of:

first difference in a ratio of object allocation and a ratio of object de-allocation of a class;

a change in the first difference over time while the program is running;

a change in the first difference after a transaction of the program is executed;

second difference in a number of allocations and a number of de-allocation for an object;

a change in the second difference over time while the program is running;

a change the second difference after a transaction of the program is executed; and

a change in a heap size of the program.

23. The system of claim 21 wherein the memory leak rate is a slope of a function of heap size over time.

24. The system of claim 21 wherein determining the number of allocations of the object is by one or a combination of:

instrumenting a class file including computer code related to the object; and

requesting an interface to provide information related to the number.