Method and system for providing checkpointing to windows application groups

• US 9,286,109 B1
• Filed: 12/15/2008
• Issued: 03/15/2016
• Est. Priority Date: 08/26/2005
• Status: Expired due to Fees

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Open Invention Network LLC
• Inventors: Backensto, Keith Richard, Havemose, Allan
• Primary Examiner(s): Onat, Umut
• Application Number:
  US12/334,660
• Time in Patent Office: 2,647 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 11/1407 Checkpointing the instruction stream
• CPC Class Codes: G06F 11/1438 Restarting or rejuvenating
• CPC Class Codes: G06F 11/1451 by selection of backup contents
• CPC Class Codes: G06F 11/1471 involving logging of persistent data for recovery
• CPC Class Codes: G06F 11/1482 by means of middleware or OS functionality
• CPC Class Codes: G06F 11/203 using migration
• CPC Class Codes: G06F 2201/805 Real-time
• CPC Class Codes: G06F 2201/84 Using snapshots, i.e. a logical point-in-time copy of the data
• CPC Class Codes: G06F 2201/85 Active fault masking without idle spares
• CPC Class Codes: G06F 9/30145 Instruction analysis, e.g. decoding, instruction word fields
• CPC Class Codes: G06F 9/461 Saving or restoring of program or task context
• CPC Class Codes: G06F 9/4843 by program, e.g. task dispatcher, supervisor, operating system
• CPC Class Codes: G06F 9/4856 resumption being on a different machine, e.g. task migration, virtual machine migration G06F9/5088 takes precedence

Abstract

A method and system of checkpointing multi-threaded applications, and multi-process application groups on WINDOWS® operating systems. In an exemplary embodiment, the method may include creating at least one full checkpoint for each application in an application group, and creating at least one incremental application checkpoint for each application in the application group. Further, each of the at least one incremental application checkpoint may be automatically merged against a corresponding full application checkpoint. Further, checkpointing may be synchronized across all applications in the application group. Further, checkpointing may be configured to perform live migration. In the exemplary embodiment, checkpoints are triggered asynchronously using Asynchronous Procedure Calls (APC).

Claims

1. A method comprising:

creating at least one full checkpoint for an application group, wherein each of said at least one full checkpoint is comprised of a full checkpoint for each application in said application group, and at least one incremental checkpoint for each application in said application group;

merging each of said at least one incremental checkpoint for each application in said application group against a corresponding application full checkpoint;

creating an application group checkpointing thread and a checkpointing thread for each of the one or more applications comprising the application group;

triggering the application group'"'"'s checkpointer thread using asynchronous procedure calls (APC), wherein the application group'"'"'s checkpointer thread further triggers the checkpointing threads for each of the one of more applications using APC;

creating a shared barrier synchronization barrier accessible by the one or more applications comprising the application group, wherein said application group checkpointer thread releases each application'"'"'s checkpointer thread from the shared barrier synchronization barrier to checkpoint each individual application; and

creating, for each application in said application group, a buffer filled with machine-executable instructions to load a checkpoint library, call the checkpoint library'"'"'s initialization function, and jump to the application'"'"'s original entry point, wherein the buffer filled with machine-executable instructions for an application is inserted into the address space of said application and the application'"'"'s entry point is changed to begin execution at the first instruction in said buffer;

wherein said application group is comprised of one or more applications, each application is comprised of one or more processes, and each process comprised of one or more threads; and

wherein checkpoints are comprised of at least one or more of thread stacks, code segments, data segments, dynamically allocated segments, heaps, APC queues, CPU registers, instruction pointer, and kernel attributes.

2. The method according to claim 1, wherein the applications in the application group are started from an initial application, and said initial application is one of the applications in the application group.

3. The method according to claim 1, wherein the applications in the application group are independently started.

4. The method according to claim 1, wherein the application group is a shell script and the applications launched by the shell script.

5. The method according to claim 1, where the application group is virtual machine software.

6. The method according to claim 1, where handles to kernel objects are included in the full and incremental checkpoints.

7. The method according to claim 1, wherein checkpointing is provided in part by a shared dynamic link library.

8. The method according to claim 1, wherein checkpointing is provided in part by a loadable kernel module.

9. The method according to claim 8, where commands are sent to the kernel module using Input/Output Control (IOCTL) commands.

10. The method according to claim 1, wherein checkpointing does not modify the individual applications in the application group.

11. The method according to claim 1, further comprising using a customized system library.

12. The method according to claim 1, wherein standard system libraries are used, and customization are performed using interception.

13. The method according to claim 1, further comprising:

a user-space exception handler to detect if a thread was interrupted to take a full or incremental checkpoint while active in kernel space before returning to user space;

setting the thread'"'"'s user-mode instruction pointer to an invalid address during a full or incremental checkpoint to trigger an exception if the thread returns to the user space after returning from a checkpoint signal handler; and

moving the thread to a next barrier from the exception handler to prepare for checkpointing the thread'"'"'s state.

14. The method according to claim 1, further comprising:

a user-space Asynchronous Procedure Calls (APC) handler to detect if a thread was interrupted to take a full or incremental checkpoint while active in kernel space before entering a wait state;

ensuring the thread'"'"'s wait state is alertable by setting its Alertable flag;

inserting a user-mode APC at the front of the thread'"'"'s APC queue during the full or the incremental checkpoint to determine if the thread enters the wait state after returning from a checkpoint signal handler; and

moving the thread to a next barrier from the user-mode APC handler to prepare for checkpointing the thread'"'"'s state.

15. The method according to claim 1, where an application from the one or more applications can join the checkpointing service for the application group at any time.

16. The method according to claim 1, where an application of the one or more applications can leave the checkpointing service for the application group at any time.

17. The method according to claim 1, further comprising restoring the application group comprised of one or more applications in its entirety from a full checkpoint of the application group;

wherein the full checkpoint of the application group is comprised of checkpoints of the one or more applications in the application group.

18. The method according to claim 17, further comprising:

recreating all kernel objects used by processes of the one or more applications in the application group; and

changing kernel object handles to reference their newly re-created kernel objects.

19. The method according to claim 17, wherein the application group is restored on a primary node.

20. The method according to claim 17, wherein the application group is restored on a backup node.

21. A method for achieving transparent integration of an application group comprised of one or more applications with a high-availability protection program, the method comprising:

injecting registration code into each application of the application group when applications are launched and registering each application of application group with the high-availability protection program;

constructing an in-memory buffer with machine executable instructions to load a checkpoint library into each newly launched application and call the checkpoint library'"'"'s initialization function directly, wherein the buffer filled with machine-executable instructions for an application is inserted into the address space of said application and the application'"'"'s entry point is changed to begin execution at the first instruction in said buffer;

creating a coordinator checkpointer thread and a checkpointer lock;

detecting a failure in execution of the application group or any of the applications within the group; and

restoring the application group and the applications within the application group from its most recent checkpoint on a backup node in response to said detecting a failure, wherein kernel attributes of each application process are restored individually, addresses of objects references an EPROCESS are changed to reflect their new memory addresses, and restoration of memory segments include one or more of thread stacks, code segments, data segments, dynamic segments, heaps, APC queues, and file mapped segments

wherein each of the one or more applications are comprised of one or more processes, and each of the one or more processes comprised of one or more threads;

wherein the coordinator'"'"'s checkpointer thread acquires the checkpointer lock and waits for each application in said application group to restore from a checkpoint before releasing said checkpointer lock.

22. The method according to claim 21, wherein the injecting step is performed without a need to modify or recompile the applications, and without a use of a custom loader.