TRANSACTION SUPPORT FOR DISTRIBUTED DATA

• US 20120179645A1
• Filed: 01/07/2011
• Published: 07/12/2012
• Est. Priority Date: 01/07/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Lomet, David, Mokbel, Mohamed, Levandoski, Justin, Zhao, Keliang
• Granted Patent: US 8,868,514 B2
• Time in Patent Office: Days
• US Class Current: 707/607
• CPC Class Codes: G06F 16/27 Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor

Abstract

A distributed system with transaction support may have a transaction component and one or more data components. The transaction component may manage a transaction using a log sequence number for each operation, and then transmit operations to one or more data components with log sequence numbers. The data components may perform the data operations in an idempotent manner and return a reply. The transaction component may then write the operation, its log sequence number, and information from the reply message to its log. The transaction component is able to commit a transaction, as well as retry or undo portions of a transaction, by using the information stored on its log. This may be possible even when a single transaction uses multiple data components, which may be located on different devices or manage separate and independent data sources.

Claims

1. A system comprising:

at least one transaction component comprising;

a processor;

a transaction component manager that;

receives a transaction comprising a plurality of operations;

for each of said operations;

determines a log sequence number for said operation;

transmits said operation to a data component with said log sequence number;

receives a reply from said data component for said operation; and

logs said operation after receiving said reply;

at least one data component comprising;

a processor;

a storage system;

a data component manager that;

receives operation requests comprising operations and a log sequence number;

processes said operation requests in an idempotent manner; and

returns a reply.

2. The system of claim 1 comprising a plurality of data components, a first data component operating a first data store and a second data component operating a second data store.

3. The system of claim 2, said transaction comprising a first operation request for a first data component and a second operation request for a second data component.

4. The system of claim 3, said transaction component manager that further:

receives a reply from said first data component for said first operation request; and

detects a problem with said second operation request and causes said first data component to undo said first operation request.

5. The system of claim 3, said transaction component manager that further:

determines an end of stable log log sequence number and transmits said end of stable log log sequence number to said first data component;

said first data component that further;

may execute all operation requests and make effects stable for operations with log sequence numbers up to said end of stable log log sequence number, said all operation requests comprising said first operation request.

6. The system of claim 3, said transaction component manager that further:

fails to receive a first reply for said first operation request; and

resends said first operation request to said first data component.

7. The system of claim 3, said transaction component manager that further:

detects a problem and determines a maximum log sequence number;

transmits said maximum log sequence number to said first data component, said first data component that undoes all operation requests with a higher log sequence number than said maximum log sequence number.

8. A method comprising:

receiving a transaction comprising a plurality of data operations comprising a first data operation for a first data component and a second data operation for a second data component, said transaction being received from a client;

for each of said data operations, obtaining a log sequence number and transmitting said data operations and said log sequence number to a data component, said log sequence number being a monotonically increasing number;

receiving an reply from each of said data components and logging said information from the reply and an associated log sequence number; and

after receiving all of said replies from said data components for all operations within a transaction, returning a transaction completion reply to said client.

9. The method of claim 8, said first data operation comprising a write operation.

10. The method of claim 9 further comprising:

creating a lock on a record associated with said first data operation prior to determining said log sequence number.

11. The method of claim 8 further comprising:

establishing an authenticated session with said client prior to receiving said transaction.

12. The method of claim 8 further comprising:

after receiving all of said replies, transmitting a first command to each of said data components, said first command permitting each of said data components to make stable said data operations to storage.

13. The method of claim 12, said first command comprising an end of stable log indicator comprising a log sequence number, said end of stable log indicator being a maximum log sequence number for which all data operations are made stable.

14. The method of claim 8 further comprising:

determining that said first data operation may not have completed; and

resending said first data operation to said first data component.

15. The method of claim 14, said first data operation having been already completed prior to said resending, said first data component performing said first data operation in an idempotent manner.

16. The method of claim 8, said receiving said reply being received out of order per said log sequence number.

17. A system comprising:

a transaction component comprising;

a processor;

a session manager operable on said processor, said session manager that establishes an authenticated session with a client process;

a record manager that receives a transaction to be performed on at least one record, said transaction being received from said client process and having a plurality of data operations comprising a first data operation performed on a first data component and a second data operation performed on a second data component;

a lock manager that obtains log sequence numbers for data operations associated with said transaction; and

a log manager that receives acknowledgments from a data component, said acknowledgments having an associated log sequence number, and logs said replies in a log file;

said record manager that transmits said data operations to data components, said data operations having said associated log sequence number.

18. The system of claim 17, said log manager that may receive said replies from multiple threads out of sequence from said log sequence number.

19. The system of claim 18, said record manager that may further:

identify an end of stable log, said end of stable log being a maximum log sequence number; and

transmits said end of stable log to said data components, said data components makes stable all operations with log sequence number less than said end of stable log.

20. The system of claim 19, said record manager that may further:

determine that said first data operation has not been confirmed complete; and

re-sends said first data operation.