NonStop Systems Introduction

Transaction Management
NonStop Systems Introduction—527825-001
5-5
Audited Files and Audit Trails
This distributed transaction is a typical TMF transaction. It begins with a BEGIN
TRANSACTION statement, performs reads and writes at multiple nodes, and
concludes with an END TRANSACTION statement. This bracketed unit of work is
complete when the END TRANSACTION statement is executed successfully by the
program. All of this work survives in the database, secured against a subsequent
hardware failure, because of TMF and other hardware and software features of
NonStop servers.
TMF can service two kinds of distributed transactions: (1) homogeneous distributed
transactions, where portions of a transaction are processed by TMF on two or more
Expand nodes, and (2) heterogeneous distributed transactions, where portions of a
transaction are processed under the control of multiple transaction managers: TMF
and one or more foreign transaction managers.
Audited Files and Audit Trails
How does TMF keep track of the work being done by a transaction? The transaction
leaves no record of its activities except for the database records it has changed. If you
had to reconstruct what a transaction did, the updated records would not be much help
because some other transaction might have come along and changed the records
again before you got a chance to look at them. So you need some other mechanism.
As you can see from Figure 5-3 on page 5-6, TMF uses the mechanism of an audit
trail to monitor all the changes made by a transaction. You just need to designate the
database files that are used by your application as audited files (files protected by TMF
during transactions). Then each update to one of these files by a TMF transaction
generates audit trail information.