Data Synchronization Method and Apparatus

• US 20140122934A1
• Filed: 12/30/2013
• Published: 05/01/2014
• Est. Priority Date: 08/22/2011
• Status: Active Grant

• Current Assignee: Huawei Technologies Co., Ltd. (Huawei Investment & Holding Co., Ltd.)
• Original Assignee: Huawei Technologies Co., Ltd. (Huawei Investment & Holding Co., Ltd.)
• Inventors: Wang, Zhenguo
• Granted Patent: US 9,483,382 B2
• Time in Patent Office: Days
• US Class Current: 714/37
• 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: G06F 11/36 Preventing errors by testing or debugging software
• CPC Class Codes: G06F 11/3636 by tracing the execution of the program
• CPC Class Codes: G06F 12/0815 Cache consistency protocols
• CPC Class Codes: G06F 8/65 Updates security arrangements therefor G06F21/57
• CPC Class Codes: G06F 9/30087 Synchronisation or serialisation instructions
• CPC Class Codes: G06F 9/30145 Instruction analysis, e.g. decoding, instruction word fields
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects

Abstract

Embodiments of the present invention provide a data synchronization method and apparatus. The method includes: scanning a code to capture a synchronization instruction; replacing the captured synchronization instruction with a trap instruction; and when the code runs to the trap instruction, trapping a code execution right into a distributed shared memory (DSM) unit, where the DSM unit implements a concurrent multi-write protocol, but implements a single-write protocol when executing the synchronization instruction, thereby supporting the concurrent multi-write protocol and solving the synchronization problem caused by the concurrent multi-write protocol.

Claims

1. A data synchronization method, comprising:

scanning, by a code scanning unit, a code to capture a synchronization instruction;

replacing, by the code scanning unit, the captured synchronization instruction with a trap instruction; and

when the code runs to the trap instruction, obtaining, by a distributed shared memory (DSM) unit, a code execution right, wherein the DSM unit implements a concurrent multi-write protocol, and wherein the DSM unit implements a single-write protocol when executing the synchronization instruction.

2. The method according to claim 1, wherein before the code runs to the trap instruction, the method further comprises:

scanning, by the code scanning unit, the code to capture a flow change instruction; and

replacing the flow change instruction with the trap instruction.

3. The method according to claim 2, further comprising:

determining, by the DSM unit, whether a target page to which a flow change is directed is a scanned page when executing the flow change instruction;

when the target page to which the flow change is directed is the scanned page, executing the flow change to a corresponding code on the scanned page, and executing the corresponding code; and

when the target page to which the flow change is directed is an unscanned page, executing, for the unscanned page, the process of scanning a code to capture a synchronization instruction and the process of replacing the captured synchronization instruction with a trap instruction.

4. The method according to claim 1, wherein the synchronization instruction is a memory barrier instruction, and wherein implementing, by the DSM unit, the single-write protocol when executing the synchronization instruction comprises returning, by the DSM unit, the code execution right after all data update operations are completed.

5. The method according to claim 1, wherein the synchronization instruction is an atomic operation instruction, and wherein implementing, by the DSM unit, the single-write protocol when executing the synchronization instruction comprises:

broadcasting a notification to other nodes such that the other nodes set corresponding data to read-only and forbid synchronization operations for the corresponding data; and

broadcasting updated data to the other nodes after execution of the atomic operation instruction is completed such that the other nodes update the corresponding data and allow synchronization operations for the corresponding data.

6. The method according to claim 1, further comprising obtaining, by the DSM unit, the code execution right when a page fault occurs in a running process of the code.

7. The method according to claim 6, further comprising:

when the page fault is caused by a read operation, generating, by the DSM unit, a copy of data to which the read operation is directed, and causing other nodes to set corresponding copies on the other nodes to read-only;

when the page fault is caused by a write operation and a page to which the write operation is directed does not exist, requesting, by the DSM unit, a cache, and writing data to the cache; and

when the page fault is caused by the write operation and a page to which the write operation is directed exists, updating, by the DSM unit, data, and broadcasting the updated data and updated global time to the other nodes such that the other nodes update corresponding data according to the global time.

8. The method according to claim 1, wherein the concurrent multi-write protocol is a write-update protocol.

9. A data synchronization apparatus, comprising:

a code scanning unit configured to scan a code to capture a synchronization instruction, and replace the captured synchronization instruction with a trap instruction; and

a distributed shared memory (DSM) unit configured to obtain a code execution right when the code runs to the trap instruction, wherein the DSM unit implements a concurrent multi-write protocol, and wherein the DSM unit implements a single-write protocol when executing the synchronization instruction.

10. The apparatus according to claim 9, wherein the code scanning unit is further configured to scan the code to capture a flow change instruction, and replace the flow change instruction with the trap instruction.

11. The apparatus according to claim 10, wherein the DSM unit is further configured to:

when executing the flow change instruction, determine whether a target page to which a flow change is directed is a scanned page;

when the target page to which the flow change is directed is the scanned page, execute the flow change to a corresponding code on the scanned page, and execute the corresponding code; and

when the target page to which the flow change is directed is an unscanned page, call the code scanning unit to execute, for the unscanned page, the process of scanning a code to capture a synchronization instruction and the process of replacing the captured synchronization instruction with a trap instruction.

12. The apparatus according to claim 9, wherein the synchronization instruction is a memory barrier instruction, and wherein the DSM unit is further configured to return the code execution right after all data update operations are completed.

13. The apparatus according to claim 9, wherein the synchronization instruction is an atomic operation instruction, and wherein the DSM unit is further configured to:

broadcast a notification to other nodes such that the other nodes set corresponding data to read-only and forbid synchronization operations for the corresponding data; and

broadcast updated data to the other nodes after execution of the atomic operation instruction is completed such that the other nodes update the corresponding data and allow synchronization operations for the corresponding data.

14. The apparatus according to claim 9, wherein the DSM unit is further configured to obtain the code execution right when a page fault occurs in a running process of the code.

15. The apparatus according to claim 14, wherein the DSM unit is further configured to:

when the page fault is caused by a read operation, generate a copy of data to which the read operation is directed, and cause other nodes to set corresponding copies on the other nodes to read-only;

when the page fault is caused by a write operation and a page to which the write operation is directed does not exist, request a cache, and write data to the cache; and

when the page fault is caused by the write operation and the page to which the write operation is directed exists, update data, and broadcast the updated data and updated global time to the other nodes such that the other nodes update corresponding data according to the global time.

16. The apparatus according to claim 9, wherein the apparatus runs in an operating system or an application program or is comprised in an operating system.

17. The apparatus according to claim 9, wherein the concurrent multi-write protocol is a write-update protocol.

18. A data processing system, comprising:

an operating system; and

a data synchronization apparatus that runs under the operating system,

wherein the data synchronization apparatus comprises;

a code scanning unit configured to scan a code to capture a synchronization instruction, and replace the captured synchronization instruction with a trap instruction; and

a distributed shared memory (DSM) unit configured to obtain a code execution right when the code runs to the trap instruction, wherein the DSM unit implements a concurrent multi-write protocol, and wherein the DSM unit implements a single-write protocol when executing the synchronization instruction.