Extension of two phase commit protocol to distributed participants

• US 5,546,582 A
• Filed: 02/02/1995
• Issued: 08/13/1996
• Est. Priority Date: 08/20/1992
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Dievendorff, Richard, LaBelle, Margaret K., House, Daniel E., Jenner, Earle H., Silen, Stuart L., Mall, Michael G., Brockmeyer, Roger L.
• Primary Examiner(s): Lim, Krisna
• Assistant Examiner(s): Coulter, Kenneth R.
• Application Number:
  US08/384,484
• Time in Patent Office: 558 Days
• Field of Search: 395/200, 395/200.19, 395/600, 395/650
• US Class Current: 709/201
• CPC Class Codes: G06F 16/2379 Updates performed during online database operations; commit processing

Abstract

An extension of the two phase commit protocol allows distributed participation among physically distant agents independent of the communications mechanism being used in a data processing system. An extra stage of processing is added to the two phase commit protocol called End Phase One Processing (EPOP) which enables a distribution of the coordinator function across systems using any communication mechanism. EPOP is an extra stage in which a participant can receive control. In this extra stage, a participant flows two phase commit protocol sequences to distributed systems. The communication mechanism is used in such a way that it becomes part of a distributed coordinator. The coordinator itself does not need knowledge of other systems. The extra stage of processing is enabled by an operating system service called Enable End Phase One Exit Processing (EEPOEP). EEPOEP causes an extension of two phase commit protocol to be used on the issuing system. In this way, not only distributed databases can be supported, but also distributed users and distributed generic resource managers. A new response, called ABSTAIN, can be used by a resource manager in response to a PREPARE signal from the coordinator. This response from the resource manager to the coordinator indicates that the resource manager wants to continue to be involved with the unit of work two phase commit process but does not want to influence the final decision (i.e., COMMIT or BACKOUT) of the unit of work.

Claims

1. A method for coordinating updates in a distributed data processing system comprising the steps of:

notifying, from an external source, a first coordinator on a first system in said distributed data processing system so as to enable said first coordinator to, at a predetermined time, invoke an end-phase-one exit processing, said end-phase-one exit processing comprising a function provided by said external source for performing processing after an end of a phase-one of a plurality of phases;

enabling said end-phase-one exit processing of said first coordinator so that said first coordinator on said first system becomes a sub-coordinator of a second coordinator on a second system in said distributed data processing system, wherein said sub-coordinator notifies participants to perform said phase-one, collects votes from participants within said first system and writes an end-phase-one record of a first system vote to a log when all of said votes are collected;

notifying, by said first coordinator, participants to perform said phase-one;

collecting, by said first coordinator, all votes from all participants on said first system in said distributed data processing system;

writing, by said first coordinator, said end-phase-one record of said first system vote to said log when all votes are collected from all participants on said first system; and

in response to said writing by said first coordinator of said first system, passing control, by said first coordinator, to said end-phase-one exit processing,

wherein said votes include an ABSTAIN vote when a participant wants to continue to be involved with a unit of work but does not want to influence a final decision of the unit of work.

2. The method for coordinating updates in a distributed data processing system recited in claim 1 wherein said votes to prepare include at least one of YES, indicating a participant is prepared to COMMIT changes to a unit of work, NO, indicating a participant wants to BACKOUT of changes to a unit of work, and ABSTAIN, indicating a participant wants to continue to be involved with the unit of work but does not want to influence the final decision of the unit of work.

3. The method according to claim 1, wherein said first coordinator and said second coordinator are separated from said first communication manager and said second communication manager, respectively,

wherein one of said first system and said second system selectively operates as a sub-coordinator for the other of said one of said first operating system and said second operating system, said other of said one of said first operating system and said second operating system being a coordinator function.

4. The method according to claim 1, wherein said first coordinator and said second coordinator are unaware of said participants on said second system and said first system, respectively, and

wherein one of said first coordinator and said second coordinator selectively operates as a sub-coordinator for the other of said one of said first coordinator and said second coordinator.

5. The method according to claim 1, wherein said first coordinator and said second coordinator are unaware of said participants on said second system and said first system, respectively, and

wherein said first coordinator and said second coordinator are separated from said first communication manager and said second communication manager, respectively, and wherein said first communication manager and said second communication manager are for operating without knowledge of and responsibilities to coordinate a two phase commit sequence.

6. A method for coordinating updates in a distributed data processing system using a commit protocol and including at least a first machine running a first operating system and a second machine running a second operating system, said method comprising the steps of:

sending by said first operating system a prepare signal to participants on said first machine, one of said participants being a first communication manager communicating with a second communication manager on said second machine, said first communication manager sending the prepare signal to said second communication manager;

notifying, by said second communication manager, in response to said prepare signal from said first communication manager, said second operating system to enable an end-phase-one exit so that said second operating system becomes a sub-coordinator of a coordinator function of said first operating system, said sub-coordinator notifies participants to perform a phase-one of a plurality of phases, collects votes from participants on said second machine and writes an end-phase-one record of a second machine vote to a log when all of said votes are collected, said end-phase-one record representing an end of said phase-one, said end-phase-one exit comprising a function for informing said second communications manager of a result from said phase-one processing, said second communication manager informing said first communication manager of the result from said second system in response to said first system'"'"'s prepare signal;

sending by said second operating system a prepare signal to participants on said second machine;

receiving votes by said second operating system from said participants on said second machine; and

passing control from said second operating system to said second communication manager of said second operating system using the end-phase-one exit, said second communication manager sending said votes to said first communication manager,

wherein said votes include an ABSTAIN vote when a participant wants to continue to be involved with a unit of work but does not want to influence a final decision of the unit of work.

7. The method for coordinating updates in a distributed data processing system recited in claim 6 wherein said votes in response to prepare are at least one of YES, indicating a participant is prepared to COMMIT changes to a unit of work, NO, indicating a participant wants to BACKOUT of changes to a unit of work, FORGET, indicating a participant'"'"'s vote is READ ONLY, and ABSTAIN, indicating a participant wants to continue to be involved with the unit of work but does not want to influence the final decision of the unit of work.

8. The method for coordinating updates in a distributed data processing system recited in claim 7 wherein at least some of said votes from said participants on said second machine are YES,

said method further comprising the step of writing by said second operating system said end-phase-one record of a vote by said participants on said second machine to a second machine log,

said step of passing control to said second communication manager being performed using the end-phase-one exit after writing to said second machine log.

9. The method for coordinating updates in a distributed data processing system recited in claim 7 wherein all of said votes by participants on said second machine are FORGET or ABSTAIN, said method further comprising the steps of:

writing by said first operating system an atomic instant record to a first machine log upon receiving all votes from participants on said first machine, said votes including votes received by said first communication manager from said second communication manager; and

sending by said first operating system an action signal to participants on said first machine, said action signal being determined by an outcome of said votes, but no action signal being sent to said second communication manager due to said FORGET or ABSTAIN votes by said participants on said second machine.

10. The method for coordinating updates in a distributed data processing system recited in claim 7 wherein all votes to prepare by participants on said first machine are FORGET or ABSTAIN, said method further comprising the steps of:

writing by said second operating system said end-phase-one record of a vote by said participants on said second machine to a second machine log;

after writing to said second machine log, passing control to said second communication manager, said second communication manager sending said vote to said first communication manager;

writing by said first operating system an atomic instant record to a first machine log upon receiving votes received by said first communication manager from said second communication manager;

sending by said first operating system an action signal to said first communication manager but not to participants on said first machine voting FORGET or ABSTAIN, said first communication manager sending the action signal to said second communication manager;

writing by said second operating system an action record to said second machine log in response to said action signal received by said second communication manager; and

sending by said second communication manager to said first communication manager an output that an action has been completed.

11. The method for coordinating updates in a distributed data processing system recited in claim 7 further comprising the steps of:

writing by said first operating system an atomic instant record to a first machine log upon receiving all votes from participants on said first machine, said votes including votes received by said first communication manager from said second communication manager;

sending by said first operating system an action signal to participants on said first machine, said action signal being determined by an outcome of said votes, said first communication manager sending the action signal to said second communication manager;

writing by said second operating system an action record to said second machine log in response to said action signal received by said second communication manager; and

sending by said second communication manager to said first communication manager an output that an action has been completed.

12. The method for coordinating updates in a distributed data processing system recited in claim 11 further comprising the steps of:

sending by said second operating system said action signal to participants on said second machine and upon receiving acknowledgements from participants on said second machine that an action has been completed, updating said second machine log; and

said first operating system, upon receiving acknowledgements from participants on said first machine, including said first communication manager, that an action has been completed, updating said first machine log.

13. The method according to claim 6, wherein said votes from at least one of said participants and said communication manager include at least one ABSTAIN vote when said at least one of said participants and said communication manager wants to continue to be involved with a unit of work but does not want to influence a final decision of the unit of work.

14. The method according to claim 6, wherein said coordinator and said subcoordinator are separated from said first communication manager and said second communication manager, respectively,

one of said first operating system and said second operating system selectively operating as one of a sub-coordinator for the other of said one of said first operating system and said second operating system, said first operating system being unaware of participants on said second machine and said second operating system being unaware of participants on said first machine,

said coordinator and said subcoordinator being separated from said first communication manager and said second communication manager, respectively,

wherein said first communication manager and said second communication manager are for operating without knowledge of and responsibilities to said extended two phase commit protocol.

15. The method for coordinating updates in a distributed data processing system recited in claim 7 wherein at least one of said votes by participants on said second machine is BACKOUT,

said method further comprising the steps of:

writing by said first operating system an atomic instant record to a first machine log upon receiving all votes from participants on said first machine, said votes including votes received by said first communication manager from said second communication manager.

16. The method for coordinating updates in a distributed data processing system recited in claim 7, wherein at least one vote to prepare by participants on said first machine is BACKOUT,

said method further comprising the steps of:

writing by said second operating system an end-phase-one record of a vote by said participants on said second machine to a second machine log;

after writing to said second machine log, passing control to said second communication manager, said second communication manager sending said vote to said first communication manager;

sending by said first operating system an action signal to said first communication manager but not to participants on said first machine voting FORGET or ABSTAIN, said first communication manager sending the action signal to said second communication manager; and

sending by said second communication manager to said first communication manager an output that an action has been completed.

17. A distributed data processing system which implements an extended two phase commit protocol including a first phase representing a phase-one of said commit protocol, said system comprising:

at least a first machine running a first operating system, said first operating system supporting a first plurality of participants including a first communication manager;

at least a second machine running a second operating system, said second operating system supporting a second plurality of participants including a second communication manager; and

communication means for interconnecting said first and second communication managers,

said first operating system sending a prepare signal to said first plurality of participants on said first machine and said first communication manager sending the prepare signal to said second communication manager,

said second operating system, in response to a prepare signal from said first communication manager, notifying said second communication manager to enable an end-phase-one exit so that said second operating system becomes a subcoordinator of a coordinator function of said first operating system, said subcoordinator notifies participants to perform said phase-one, collects votes from participants on said second operating system and writes an end-phase-one record of said second operating system vote to a log when all said votes are collected, said end-phase-one exit comprising a function for informing said second communication manager that said second operating system is said sub-coordinator of said coordinator function of said first operating system and informing said second communication manager of an outcome of said votes from participants on said second operating system,

said second operating system sending a prepare signal to participants on said second machine,

said second operating system writing said end-phase-one record of a vote by said participants on said second machine to a second machine log, and after writing to said second machine log, passing control to said second communication manager using the end-phase-one exit,

said second communication manager sending said vote to said first communication manager,

wherein said votes include an ABSTAIN vote when a participant wants to continue to be involved with a unit of work but does not want to influence a final decision of the unit of work.

18. The distributed data processing system recited in claim 17 wherein said first and second machines are virtual machines running on a single computer under a virtual machine operating system, said first and second operating systems being guest operating systems of said virtual machine operating system.

19. The distributed data processing system recited in claim 17 wherein said first and second machines are implemented on physically separate computers and said communication device includes a communication link interconnecting said physically separate computers.

20. The system according to claim 17, wherein said system includes means for issuing an ABSTAIN vote by any of said participants when a participant of said any of said participants wants to continue to be involved with a unit of work but does not want to influence a final decision of the unit of work.

21. The system according to claim 17, wherein said coordinator and said subcoordinator are separated from said first communication manager and said second communication manager, respectively,

said first operating system being unaware of participants on said second machine and said second operating system being unaware of participants on said first machine,

wherein said first communication manager and said second communication manager are for operating without knowledge of and responsibilities to said extended two phase commit protocol.