Systems and methods for transaction chaining

Abstract

A transaction management engine, such as a business process management (BPM) engine, can allow an application to define transaction demarcations in order to ensure that portions of a workflow are processed atomically. In one such system, a JMS message is queued to the workflow as part of an existing transaction. If that transaction commits, the queued JMS message ensures that the workflow will be invoked in a new transaction that allows the workflow to perform the next unit of work in the newly initiated transaction. When the unit of work is completed, the workflow queues another JMS message. If this transaction commits, the transaction ensures the unit of work has executed in a transaction. When the workflow is invoked again due to the second JMS message, the workflow can continue its operation in a different transaction.

Description

COPYRIGHT NOTICE

[0002] A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document of the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.

CROSS-REFERENCED CASES

[0003] The following applications are cross-referenced and incorporated herein by reference:

[0004] U.S. Provisional Patent Application No. 60/376,906 entitled “COLLABORATIVE BUSINESS PLUG-IN FRAMEWORK,” by Mike Blevins, filed May 1, 2002;

[0005] U.S. Provisional Patent Application No. 60/377,157 entitled “SYSTEM AND METHOD FOR COLLABORATIVE BUSINESS PLUG-INS” by Mike Blevins, filed May 1, 2002.

[0006] U.S. Provisional Patent Application No. 60/450,074 entitled “SYSTEMS AND METHODS UTILIZING A WORKFLOW DEFINITION LANGUAGE” by Pal Takacsi-Nagy, filed Feb. 25, 2003.

FIELD OF THE INVENTION

[0007] The present invention relates to the filtering of messages in a subscription-based messaging model.

BACKGROUND

[0008] In a normal transaction management approach, a first transaction for a first application is “followed” by a second transaction for either the first application or for a second application. Any resources that the first application is holding against a database or resource manager are deleted when the first transaction completes. Further, any context for the first transaction is deleted upon completion. When the second transaction begins, all resources for that transaction must be re-established, including any resources used for the first transaction. This can be inefficient if any of those resources are being used by multiple transactions.

BRIEF SUMMARY

[0009] Systems and methods in accordance with one embodiment of the present invention overcome deficiencies in existing transaction management systems by taking advantage of transaction chaining. A transaction management engine, such as a business process management (BPM) engine, can allow an application to define transaction demarcations in order to ensure that portions of a workflow are processed atomically.

[0010] Other features, aspects, and objects of the invention can be obtained from a review of the specification, the figures, and the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

[0011] FIG. 1 is a diagram showing a workflow that can be used in accordance with one embodiment of the present invention.

[0012] FIG. 2 is a diagram of another workflow that can be used in accordance with one embodiment of the present invention.

[0013] FIG. 3 is a code example showing transaction demarcations within a workflow that can be used with the diagram of FIG. 2.

[0014] FIG. 4 is a chart showing an execution path that can be used with the diagram of FIG. 2.

DETAILED DESCRIPTION

[0015] Transaction managers can be used to provide atomicity for a transaction. The “atomicity” of a transaction refers to the fact that a transaction should be done as a single “atomic” unit of work, such that the entire transaction either commits or fails. If a transaction includes multiple updates to a database table, for example, a transaction manager can be used to ensure that either all of the updates get committed to the database, or that none of the updates get committed. Ensuring atomicity through a transaction manager can be substantially more complicated, however, as a transaction manager can be responsible for managing multiple, distributed transactions on a number of databases distributed throughout a system or across a network or server cluster.

[0016] Systems and methods in accordance with the present invention provide a transaction manager with the ability to manage a “chained” transaction, or a transaction chain. When allowing for a transaction chain, a transaction manager can maintain certain resources, or can maintain a transaction context, for subsequent transactions. A transaction management engine, such as a business process management (BPM) engine, can allow an application to define transaction demarcations. For example, two tasks can be completed in a business process that transfers money between accounts. First, a specific amount of money needs to be deducted from a first account. Second, that money needs to be transferred into a second account. Assuming that these are separate items of work, it can be desirable to bracket these two activities in a transaction to be an atomic work item, such that either both work items process or neither of them process. It would be undesirable to deduct money from the first account if the money did not actually get added to the second account, and it would not be desirable to add money to the second account if that money was not actually taken out of the first account.

[0017] Systems and methods in accordance with one embodiment of the present invention utilize an entity bean to implement this BPM, which can take advantage of transaction bracketing. By utilizing an entity bean, transactions can be managed by an EJB container. When demarcating a transaction boundary for application activities within BPM, the transaction can simply continue to process normally until a transaction bracket is encountered. When a bracket is encountered, a persistent message is queued to the system, such as a JMS message. The queued JMS message allows the system to store context for the transaction. This allows the system to know that, upon the next transaction invoke, the system should start with the next activity within the transaction boundary. In other words, the system is processing the first transaction, then queuing a context that can be picked up and used by the next transaction.

[0018] An inbound message bean, or message-drive bean (MDB), can be created if this context is persisted through JMS. The MDB can pull this JMS message from the queue and invoke the BPM with the appropriate message. That message can include information about the next activity that to be executed as part of the transaction, allowing the transaction to maintain context.

[0019] If a J2EE-compliant system is being utilized, a mechanism such as a J2EE API could be used for suspending and resuming transactions. When encountering a transaction bracket, the system could suspend the first transaction, process those tasks included within the brackets in the new transaction, commit the new transaction, and resume the original transaction. A problem with such an approach exists, however, as the original transaction can fail after being resumed. The system has then processed an activity beyond the execution path, with no way to automatically undo or rollback the successfully committed transaction when the original transaction rolls back.

[0020] In an example shown in FIG. 1, there are three activities 104, or operations, to be performed under a single transaction. Since all three activities are under a single transaction, if the execution of any activity fails for any reason, the transaction rolls back 112 and begins again at the first activity. If the operations succeed 106, the transaction can be committed 108. This can be guaranteed by the transaction semantics. Using a system in accordance with the present invention, activities two and three can be bracketed to be done atomically, or as a single unit of work. In order to bracket activities two and three, the user can place an opening transaction tag before activity two and a closing transaction tag after activity three. These tags can indicate that the second and third activities should be done atomically, such that both activities succeed or neither succeeds.

[0021] When a message is queued as part of the transaction, the transaction can be paused to start a new transaction, in order to implement a new transaction block. If the new transaction fails, since the JMS message is in the first queue, the JMS message can be resent. This can ensure that the transaction manager always continues from the correct location. Since the JMS message is queued, an object is only sent to get that message if the transaction commits. If the queued JMS message and the transaction roll back, the system is basically one step behind and can again attempt to go forward along the path of execution. An example of one such execution path is given in FIG. 4.

[0022] A transaction manager can be configured to control the number of retries, or the period of time for retrying a transaction. Attributes in the workflow can be used to specify, for example, a retry count. The transaction itself can also have a retry count. If a JMS message is queued and the message does not commit, the system can generate a callback due to the failure, and that message can have a retry count. So even if the first attempt failed, the user can indicate how many times the message should be retried. These transactions can be implemented as an annotation of a Java Web service (JWS) file in the workflow language (WFL).

[0023] When a workflow is implemented in J2EE as an entity bean with container-managed transactions, it can be desirable to have a mechanism to demarcate parts of the workflow as atomic units of work. In other words, it is desirable to commit the existing container-initiated transaction and start a new transaction in order to perform the next unit of work. An example of such a workflow is shown in FIG. 2, where once the workflow is started 200, a message is received 202 to execute the first transaction 204, followed by a message 206 to execute the second transaction 208. After executing the transactions, or failing to execute the transactions, a reply 210 can be sent that can indicate an end 212 to the workflow. Committing the existing transaction and having the ability to do the next unit of work in a new transaction can be treated as a single operation.

[0024] In order to achieve atomicity, a JMS message can be queued to the workflow as part of the existing transaction. If the existing transaction commits, the transaction ensures that the workflow will be invoked, due to the queued JMS message, in a different transaction that can allow the workflow to perform the next unit of work in the newly initiated transaction. When the unit of work is completed, the workflow can queue another JMS message. If this transaction commits, the transaction can ensure that the unit of work has executed in a transaction. When the workflow is invoked again, due to the second JMS message, the workflow can continue its operation in a different transaction.

[0025] A transaction can have a tag called <transaction>. A user can include tasks to be performed atomically within the transaction tags. The implication is that all the tasks to be performed should be executed under a single transaction. For example, consider the following workflow:

<process>
<invoke someMethod/>
<transaction>
<invoke method1>
<invoke method2>
</transaction>
<invoke someOtherMethod/>
</process>

[0026] In the above example, someMethod will be executed in a first transaction. The methods method1 and method2 will be executed in a second transaction, demarcated by the opening and closing <transaction> tags. Yet another transaction can be used to execute someOtherMethod. When the workflow encounters a <transaction> tag, the workflow can queue a JMS message containing the information as to the next activity that needs to be executed. Queuing a JMS message as part of an existing transaction also ensures that the workflow will continue only if that transaction commits. The next unit of work is then “chained” to the previous work in the process.

[0027] The <transaction> tags ensure that method1 and method2 can be executed atomically. Without such a mechanism, there is no way to demarcate method1 and method2 as a single unit of work using J2EE transactional APIs. It is not possible to simply suspend the current transaction, start a new transaction to do the unit of work, and resume the original transaction, as there will be problems if the new transaction commits and the original transaction rolls back.

[0028] Transaction blocks not only can ensure that all operations within a transaction block are performed atomically, but a transaction block can also provide a timeout value for the operations to complete. A transaction block can also provide for retrying the operations in case of failures or timeouts. For example, as shown in FIG. 3, a user can set separate timeout and retry values for each transaction within a workflow. In the first transaction, the timeout value is set to 30 seconds with a retry count of 3, while in the second transaction the timeout value is 20 seconds and the retry count is 2.

[0029] Atomic Transactions

[0030] Atomic transactions are known as “all-or-nothing” transactions. Actions taken prior to committing an atomic transaction are considered to be tentative or pending, and not persistent or visible to other activities. When an application finishes, the application can make a request to a coordination component to determine an outcome for the transaction. The coordination component can determine whether any of the units of work failed, such as by asking the processes doing those units of work to “vote” or reply whether or not the work items processed successfully. If all the processes vote indicate successful execution, the coordination component can commit all the actions taken. If a process indicates an aborted execution, or does not respond to the request, the coordination component can abort all actions and roll back the transaction. A commit operation makes the tentative actions persistent and visible to other transactions, while a rollback makes the tentative actions appear as if the actions never happened. Atomic transactions can provide consistent failure and recovery semantics, so applications do not need to deal with the mechanics of determining a mutually agreed outcome decision or to figure out how to recover from a large number of possible inconsistent states.

[0031] The foregoing description of preferred embodiments of the present invention has been provided for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations will be apparent to one of ordinary skill in the art. The embodiments were chosen and described in order to best explain the principles of the invention and its practical application, thereby enabling others skilled in the art to understand the invention for various embodiments and with various modifications that are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the following claims and their equivalence.

Claims

1. A method for ensuring atomicity in a workflow, comprising: demarcating tasks in a workflow to be processed atomically.

2. A method according to claim 1, further comprising: committing the tasks if all tasks to be processed atomically are successfully processed.

3. A method according to claim 1, further comprising: aborting the tasks if any of the tasks to be processed atomically are not successfully processed.

4. A method for ensuring atomicity in a workflow, comprising: queuing a message to a workflow when starting an atomic unit of work in a first transaction; invoking the workflow in a second transaction if the first transaction commits, the message causing the workflow to be invoked; allowing the workflow to perform the next unit of work in the second transaction; and queuing another message such that when the second transaction commits and the workflow is subsequently invoked, the workflow can continue its operation in a new transaction.

5. A computer-readable medium, comprising: means for demarcating tasks in a workflow to be processed atomically; and means for atomically processing those tasks.

6. A computer program product for execution by a server computer for ensuring atomicity in a workflow, comprising: computer code for demarcating tasks in a workflow to be processed atomically; and computer code for atomically processing those tasks.

7. A system for ensuring atomicity in a workflow, comprising: means for demarcating tasks in a workflow to be processed atomically; and means for atomically processing those tasks.

8. A computer system comprising: a processor; object code executed by said processor, said object code configured to: demarcate tasks in a workflow to be processed atomically; and atomically process those tasks.

9. A computer data signal embodied in a transmission medium, comprising: a code segment including instructions to demarcate tasks in a workflow to be processed atomically; and a code segment including instructions to atomically process those tasks.