Introduction › Distributed Processing Concepts and Terminology › Transaction Processing

Transaction Processing

Transaction processing is the set of programmatic controls that applications use to mark the start and end of code implementing a unit of work. Transactions are used to synchronize concurrent access to one or more shared resources. If a DPS, the shared resource is a database. A CA Gen DPS operates as a traditional database transaction application.

A DPS transforms data from one consistent state to another such that the modifications are considered to be Atomic, Consistent, Isolated, and Durable. Collectively they are known as the ACID properties and are described as follows:

• Atomic—Either all operations that are part of the unit of work that is associated with a transaction complete or none are performed.
• Consistent—Data that are modified by the operations of a transaction is transitioned to a consistent state.
• Isolated—Even though transactions may execute concurrently, no transaction has visibility into the work in progress of another transaction.
• Durable—After a transaction completes successfully, its change survives subsequent failures.

Transactions that complete successfully commit their associated work, leaving any modified data in a consistent state. The unsuccessful transactions abort the work in which they were engaged. Aborting a transaction undoes the operations that are considered to be part of the unit of work that is associated with the failed transaction. The results of any modified data are rolled back to the consistent state that existed prior to the execution of the aborted transaction.

Typical transaction processing for a CA Gen Distributed Processing application is limited to a Server Manager that contains the DPS. A DPS defines a complete unit of work. Upon the completion of the DPS, the DPS Server Manager is synchronized. This synchronization results in a commit or abort being performed, depending on the success or failure of the processing of the DPS.