Methods and apparatus to implement parallel transactions

• US 20070198792A1
• Filed: 06/27/2006
• Published: 08/23/2007
• Est. Priority Date: 02/22/2006
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Oracle America, Inc. (Oracle Corporation)
• Inventors: Dice, David, Shavit, Nir
• Granted Patent: US 8,065,499 B2
• Time in Patent Office: Days
• US Class Current: 711/163
• CPC Class Codes: G06F 12/0806 Multiuser, multiprocessor or multiprocessing cache systems
• CPC Class Codes: G06F 12/0815 Cache consistency protocols
• CPC Class Codes: G06F 12/0893 Caches characterised by their organisation or structure
• CPC Class Codes: G06F 9/466 Transaction processing
• CPC Class Codes: G06F 9/52 Program synchronisation; Mutual exclusion, e.g. by means of semaphores
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms

Abstract

A computer system includes multiple processing threads that execute in parallel. The multiple processing threads have access to a global environment including different types of metadata enabling the processing threads to carry out simultaneous execution depending on a currently selected type of lock mode. A mode controller monitoring the processing threads initiates switching from one type of lock mode to another depending on current operating conditions such as an amount of contention amongst the multiple processing threads to modify the shared data. The mode controller can switch from one lock mode another regardless of whether any of the multiple processes are in the midst of executing a respective transaction. A most efficient lock mode can be selected to carry out the parallel transactions. In certain cases, switching of lock modes causes one or more of the processing threads to abort and retry a respective transaction according to the new mode.

Claims

1. A method comprising:

maintaining an environment to execute multiple processing threads in parallel; and

during execution of the multiple processing threads, switching a manner of the execution between transactional memory usage and non-transactional memory usage of a region of respective shared data.

2. A method as in claim 1, wherein switching the manner of execution includes:

switching between multiple types of lock modes including a centralized lock mode associated with non-transactional memory usage in which the multiple processing threads compete to obtain a coarse-grained lock over respective shared data and a distributed lock mode associated with transactional memory usage in which the multiple processing threads compete to obtain finer-grained locks over sub-portions of the respective shared data, the method further comprising;

monitoring a level of contention of the multiple processing threads competing to obtain locks and, thereafter, modify contents of the respective shared data; and

wherein switching the manner of the execution occurs in response to detecting that one of the central lock mode and the distributed lock mode is more suitable for carrying out the execution of the multiple processing threads.

3. A method as in claim 1, wherein switching the manner of the execution occurs in response to detecting which one of the transactional memory usage and the non-transactional memory usage would be a more efficient memory usage mode to carry out the execution of the multiple processing threads;

and

the method further comprising initiating a change of a currently implemented memory usage mode associated with the environment to the more efficient memory usage mode.

4. A method as in claim 1 further comprising:

maintaining globally accessible mode setting information enabling each of the multiple processing threads to identify a current memory usage mode of the environment for executing the multiple processing threads in parallel.

5. A method as in claim 4 further comprising:

enabling the multiple processing threads to identify a type of memory usage mode and selectively abort further execution.

6. A method as in claim 1 further comprising:

initiating the non-transactional memory usage to include global version information and a respective global lock bit accessible by the multiple processing threads, the global version information being updated by a respective processing thread each time the respective processing thread modifies contents of the respective shared data, the respective global lock bit enabling each of the multiple processing threads to obtain a lock over the respective shared data preventing other processing threads from modifying the contents of the shared data.

7. A method as in claim 6 further comprising:

initiating the transactional memory usage to include a distributed lock mode, the distributed lock mode including corresponding segment version information and a respective segment lock bit for each of multiple segments of the respective shared data accessible by the multiple processing threads, the corresponding segment version information being updated by a respective processing thread each time the respective processing thread modifies contents of the corresponding segment of the respective shared data, the respective segment lock bit enabling each of the multiple processing threads to obtain a lock over the corresponding segment of the respective shared data preventing other processing threads from modifying the contents of the corresponding segment.

8. A method as in claim 2, wherein switching the manner of the execution between the multiple types of lock modes initially requires that a global lock bit associated with the centralized lock mode be in an inactive state in which none of the multiple processing threads currently has a lock over the respective shared data.

9. A method as in claim 2, wherein switching the manner of the execution between the multiple types of lock modes at least in part includes performing an atomic write to a repository in order to simultaneously change:

i) mode information identifying a current lock mode setting of the environment, ii) version information normally used to indicate that contents of the respective shared data have been updated, and iii) global lock information to indicate release of a respective lock over the respective shared data.

10. A method as in claim 2 further comprising:

for the centralized lock mode, providing global version information that changes to a new respective data value each time one of the multiple processing threads modifies contents of the respective shared data, a given processing thread of the multiple processing threads accessing the global version information multiple times during execution of a respective transaction in order to identify whether any other processing threads modified the contents of the globally accessible shared data during execution of the given thread.

11. A method as in claim 1 further comprising:

implementing the non-transactional memory usage of shared data in which a given processing thread carries out steps of;

during a read phase, verifying that no other processing threads modifies contents of the respective shared data by monitoring global version information that is updated to a new value upon each modification to the respective shared data;

after the read phase, checking whether any one of the other processing threads currently has a lock on the respective shared data;

if there is a lock as indicated by the checking, then abort and re-execute a given processing thread;

if there is not a lock as indicated by the checking, initiating a respective commit phase with respect to the given processing thread, the commit phase including;

i) obtaining a respective lock on the respective shared data which prevents the other processing threads from modifying the respective shared data,

ii) initiate modification of the respective shared data in accordance with a respective transaction outcome associated with the given processing thread, and

iii) after the modification of the respective shared data, updating a respective data value associated with the global version information to indicate to the other processing threads that the respective shared data has been modified.

12. A method as in claim 11, further comprising:

implementing the transactional memory usage of respective shared data in which a given processing thread carries outs steps of;

maintaining copies of version information associated with respective segments of the shared data in a locally managed read set associated with a respective transaction carried out by the given processing thread, the read set not being accessible by the other processes using the respective shared data, the read set identifying versions associated with each of the respective segments of shared data used to generate a respective transaction outcome, the copies of version information indicating respective versions of multiple accessed segments at a time when the respective transaction retrieves respective data values associated with the multiple segments of shared data from the environment; and

after acquiring a respective lock associated with a given segments of the respective shared data, verifying that current version information associated with each of the multiple segments of shared data previously used to generate the respective transaction outcome matches the version information in the locally managed read set associated with the respective transaction.

13. A computer readable media encoded with a computer program including:

instructions to initiate execution of at least one processing thread of multiple processing threads executed in parallel; and

instructions to switch a manner of the execution of the at least one processing thread between transactional memory usage and non-transactional memory usage of a region of shared data.

14. A computer readable media as in claim 13, wherein the transactional memory usage is a multi-lock based transactional memory usage and the non-transactional memory usage is a single-lock based memory usage of the region of shared data, the computer program further including:

instructions to carry out a respective transaction in accordance with a centralized lock mode in which the multiple processing threads compete to obtain a coarse-grained lock over the region of shared data; and

instructions to carry out a respective transaction in accordance with a distributed lock mode in which the multiple processing threads compete to obtain finer-grained locks over sub-portions of the region of shared data.

15. A computer readable media as in claim 13, the computer program further comprising:

instructions to check globally accessible information indicating a current respective memory usage mode for carrying out the multiple processing threads in parallel; and

instructions to switch the manner of the execution of the at least one processing thread in response to detecting a change from one of the transactional memory usage and the non-transactional memory usage to a more suitable memory usage mode for carrying out the execution of the multiple processing threads.

16. A computer readable media as in claim 13, the computer program further comprising:

instructions to abort execution of a respective transaction associated with the at least one processing thread after identifying a change in a type of memory usage to either transactional memory usage or non-transactional memory usage.

17. A computer system including:

multiple processing threads executing in parallel; and

a global environment accessible by the multiple processing threads, the global environment including i) shared data utilized by the multiple processing threads, ii) mode setting information indicating a current lock mode for executing the multiple processing threads in parallel, the current lock mode being one of a first lock mode for implementing multi-lock based transactional memory usage and a second lock mode for implementing single-lock based memory usage of the shared data.

18. A computer system as in claim 17, wherein the first lock mode is a centralized lock mode in which the multiple processing threads compete to obtain a coarse-grained lock over the shared data;

and

wherein the second lock mode is a distributed lock mode in which the multiple processing threads compete to obtain finer-grained locks over sub-portions of the shared data.

19. A computer system as in claim 17, wherein the global environment includes:

global lock information enabling a respective one of the multiple processing threads to obtain a respective lock and prevent other processing threads from modifying any of the shared data when in the first lock mode; and

segment lock information for each of multiple segments of the region of shared data, the segment lock information enabling a respective one of the multiple processing threads to obtain a respective lock and prevent other processing threads from modifying contents of a corresponding segment of the shared data when in the second lock mode.

20. A computer system as in claim 19, wherein the global environment includes:

global version information enabling a respective one of the multiple processing threads to identify whether contents of the shared data has been modified by another processing thread because each of the multiple processing threads updates a value of the global version information upon modification of the shared data when in the first lock mode; and

segment version information for each of multiple segments of the shared data, the segment version information associated with the multiple segments enabling a respective one of the multiple processing threads to identify whether contents of any corresponding segments of the shared data have been modified by another processing thread because each of the multiple processing threads updates a value of corresponding segment version information upon modification of the corresponding segment of shared data when in the second lock mode.