System and method for providing checkpointing with precompile directives and supporting software to produce checkpoints, independent of environment constraints

• US 6,161,219 A
• Filed: 07/03/1997
• Issued: 12/12/2000
• Est. Priority Date: 07/03/1997
• Status: Expired due to Fees

• Current Assignee: University of Iowa Research Foundation
• Original Assignee: University of Iowa Research Foundation
• Inventors: Strumpen, Volker, Ramkumar, Balkrishna
• Primary Examiner(s): Hafiz, Tariq R.
• Assistant Examiner(s): INGBERG, TODD D
• Application Number:
  US08/888,013
• Time in Patent Office: 1,258 Days
• Field of Search: 395/701, 395/702, 395/703, 395/704, 395/705, 395/706, 717/11
• US Class Current: 717/130
• CPC Class Codes: G06F 11/1438 Restarting or rejuvenating

Abstract

A method for portable checkpointing comprising the steps of: pre-compiling an input source code; and outputting an associated output source code which includes support for portable checkpointing. The portability of the checkpoints allows migration between systems with different hardware, software, and operating systems.

The portable checkpoints are stored in a Universal Code Format (UCF) and are selectively activated at the next potential checkpoint marker after a MinTBC (Minimum Time Between Checkpoints) timer expires.

Claims

1. A method for portable checkpointing comprising the steps of:

pre-compiling an input source code of a selected programming language; and

outputting an associated output source code which includes support for portable checkpointing, wherein the portable checkpointing provides a first portable checkpoint, and wherein said associated output source code using the first portable checkpoint can be executed on a destination platform that is not limited by (a) programming language restrictions of the selected programming language, (b) operating system level variations and (c) processor restrictions including memory page size, firmware byte-ordering and word size.

2. The method for portable checkpointing of claim 1, comprising the further step of:

inserting markers for the support of portable checkpointing in original source code to produce the input source code.

3. The method for portable checkpointing of claim 1, comprising the further step of:

compiling the associated output source code to produce executable code.

4. The method for portable checkpointing of claim 3, comprising the further steps of:

executing the executable code; and

producing portable checkpoints.

5. The method for portable checkpointing of claim 4, wherein said step of producing includes producing the portable checkpoints at checkpoint times, and wherein consecutive checkpoint times are spaced at greater than or equal to a minimum time between checkpoints.

6. The method for portable checkpointing of claim 4, wherein said step of producing includes selective activation of the portable checkpoints at runtime.

7. The method for portable checkpointing of claim 4, wherein the portable checkpoints are in a universal checkpoint format.

8. The method for portable checkpointing of claim 4, wherein each of the portable checkpoints is a contiguous area of memory.

9. The method for portable checkpointing of claim 4, wherein each of the portable checkpoints is a shadow stack, wherein the shadow stack is a data structure that maintains global variables, dynamically allocated variables and automatic stack variables defining the state of executing code.

10. The method for portable checkpointing of claim 4, wherein the portable checkpoints are minimal overhead portable checkpoints.

11. The method for portable checkpointing of claim 3, wherein the portable checkpointing is portable across at least two compilers.

12. The method for portable checkpointing of claim 1, wherein the portable checkpointing is portable across operating systems, wherein the operating systems (OSs) include at least two of a command driven compatible OS, a DOS compatible OS, a windows environment compatible OS and a network compatible OS.

13. The method for portable checkpointing of claim 12, wherein the command driven compatible OS includes Unix™

OS implementations and variants thereof, and wherein the windows environment compatible OS includes MacIntosh™

compatible OS and Windows `95™

compatible OS.

14. The method for recovering of claim 1, wherein the portable checkpoints are in a universal checkpoint format, wherein the universal checkpoint format can be adapted to the selected OS and the selected processor.

15. The method for portable checkpointing of claim 1, wherein the first portable checkpoint defines a state of an executing process of the output source code.

16. The method for portable checkpointing of claim 1, wherein the input source code includes a general-purpose programming language code.

17. The method for portable checkpointing of claim 1, wherein the pre-compiling step can compile one of C code, C++ code, Fortran code, and Java™

code.

18. The method for portable checkpointing of claim 1, wherein the support for portable checkpointing includes padding for data compatibility.

19. The method for portable checkpointing of claim 1, wherein the support for portable checkpointing includes jump tables at the entry of functions.

20. The method for portable checkpointing of claim 1, wherein the support for portable checkpointing includes function call wrappers in functions.

21. The method for portable checkpointing of claim 1, wherein the input source code includes call-by-reference.

22. The method for portable checkpointing of claim 1, wherein the pre-compiling and outputting steps can output source executable on two of a RISC compatible processor, an x86-family compatible processor, an alpha-compatible processor, and a multiprocessing compatible processor hardware platform.

23. An system for portable checkpointing, comprising:

a marker inserter for receiving a selected language source code, and inserting markers for support of portable checkpointing to produce input source code; and

a pre-compiler, coupled to said marker inserter, for pre-compiling the input source code and outputting an associated output source code that includes the support of portable checkpointing, wherein the support for portable checkpointing in the associated output source code is selectively executable on each of a RISC compatible processor, an x86-family compatible processor, an alpha-compatible processor, and a multiprocessing compatible processor hardware platform.

24. The apparatus for portable checkpointing of claim 23, further comprising:

a compiler, coupled to said pre-compiler, for compiling the associated output source code to produce executable code.

25. The apparatus for portable checkpointing of claim 23, wherein the portable checkpointing is portable across at least two compilers.

26. The apparatus for portable checkpointing of claim 23, wherein the portable checkpointing is portable across operating systems.

27. The apparatus for portable checkpointing of claim 23, wherein the portable checkpointing is portable across hardware platforms.

28. The apparatus for portable checkpointing of claim 23, wherein the portable checkpointing is in a universal checkpoint format.

29. The apparatus for portable checkpointing of claim 23, wherein the portable checkpointing uses a contiguous area of memory.

30. The apparatus for portable checkpointing of claim 23, wherein the portable checkpointing uses a shadow stack.

31. The method for portable checkpointing of claim 23, wherein multiplatforms include various off-the-shelf operating system and off-the-shelf processor combinations.

32. An apparatus for general purpose checkpointing, comprising:

a pre-compiler for pre-compiling an input source code that is independent of application semantics and outputting an associated output source code which includes support of portable checkpointing and the generation of portable checkpoints, wherein the associated output source code that includes support for portable checkpointing can regenerate memory that was dynamically-allocated during execution.

33. The apparatus for portable checkpointing of claim 32, wherein the associated output source code that includes the support of portable checkpointing is processor independent and can selectively compile with no modification to executable code of the input source code and execute on two processors that are binary-incompatible.

34. The method for portable checkpointing of claim 33, wherein binary-incompatible features of byte-ordering, word-ordering and word size.

35. A method for selectively activating portable checkpointing in output source code which includes potential checkpoint markers, comprising the steps of:

selectively activating, at run time, portable checkpointing at the next potential checkpoint marker; and

producing an associated portable checkpoint in the output source code that is independent of operating system restrictions and processor restrictions including one of memory page size, firmware byte-ordering and word size for a destination operating system and processor that will execute the portable checkpoint.

36. The method of claim 35, comprising the further step of:

saving the portable checkpoint to a stable storage, wherein the associated portable checkpoint is capable of regenerating dynamically-allocated memory.

37. The method of claim 36, comprising the further step of:

waiting a minimum time between checkpoints before repeating said selectively activating step.

38. The method of claim 35, comprising the further step of:

waiting a minimum time between checkpoints before repeating said selectively activating step.

39. The method of claim 35, wherein said waiting step is executed after the completion of said saving step.

40. The method of claim 35, wherein said waiting step is executed concurrently with said saving step.

41. The method of claim 40, wherein said saving step includes latency hiding.

42. A method for checkpoint recovery using portable checkpoints, comprising the steps of:

performing execution of an executable code with the portable checkpoints in a first operating system and processor platform;

restarting the executable code on a second operating system and processor platform;

loading a first portable checkpoint taken on the first operating system and processor platform onto the second operating system and processor platform; and

resuming execution of the executable code on the second operating system and processor platform from a point in the executable code where the first portable checkpoint was taken, wherein the first portable checkpoint is independent of the first and second platforms executing the executable code.

43. The method for recovering of claim 42, wherein the portable checkpoints are fault tolerant portable checkpoints.

44. The method for recovering of claim 42, wherein the portable checkpoints are distributed processing portable checkpoints.

45. The method for recovering of claim 42, wherein the portable checkpoints are load balancing portable checkpoints.

46. The method for recovering of claim 42, wherein the portable checkpoints are suspension and subsequent execution portable checkpoints.

47. The method for recovering of claim 42, wherein the portable checkpoints are diagnostic portable checkpoints.

48. The method for recovering of claim 42, wherein said step of loading includes loading the recovery portable checkpoint into a shadow stack.

49. The method for recovering of claim 42, comprising the further step of:

restoring at least one of data/bss, heap segment, and stack.

50. The method for recovering of claim 42, wherein the recovery portable checkpoint is a last portable checkpoint.

51. The method of claim 42, wherein the OS is selectable from each of a command driven compatible OS, a DOS compatible OS, a windows environment compatible OS and a network compatible OS, and wherein the processor is selectable from all of a RISC compatible processor, an x86-family compatible processor, an alpha-compatible processor, and a multiprocessing compatible processor hardware platform.