TS/MP 2.5 Pathsend and Server Programming Manual

Data Integrity
If your database is corrupted by a hardware or software failure, you might need weeks to isolate
and then correct the problem. Because an inaccessible or inconsistent database can have a
dramatic, adverse effect on business operations, HP developed the TMF subsystem, to ensure
database consistency. The TMF subsystem, which works with TS/MP, protects the entire database
from catastrophic system failures by maintaining an audit trail of database changes (that is,
transactions); an audit trail is also commonly known as a transaction log. You can use the audit
trail to rebuild the database in the event of a hardware or software failure.
The design of Pathway servers supports the integrity of individual transactions and therefore
transaction processing protection as a whole. Because the requestor/server model allows a clear
division of processing functions, application programmers can code each server program to handle
a specific set of transaction types: for example, checking an account balance, entering a new
customer, or updating the parts inventory. The server processes service their transactions by
performing the same set of tasks over and over again. In this way, a valid transaction is defined
as a specific set of tasks both by the requestor program and within the server logic.
If for any reason a server is unable to complete all tasks involved in processing a transaction, it
can abort the transaction and thereby maintain the transaction’s integrity. The server does not have
to wait for the requestor to abort the transaction.
Fault Tolerance
Because OLTP systems automate core business operations and deliver key business services,
companies depend on OLTP applications to stay up and running—even if a hardware or software
component fails.
NonStop systems, which are specifically intended for online transaction processing, are designed
to remain continuously available during the hours when transactions are being entered and business
is being conducted. Typically, a NonStop system can continue processing despite the failure of
any single software or hardware component within that system. This ability is referred to as fault
tolerance.
In the Pathway environment, automatic fault tolerance (that is, fault tolerance that does not require
any additional programming effort on your part) is provided by the use of process pairs and the
actions of the PATHMON process, the TMF subsystem, and the TCP provided with the Pathway/iTS
product.
In the Guardian operating environment, the functions and tasks of an application are performed
by processes, which are running programs. A process pair consists of a primary process, which
does some specific function in the overall work of the application, and a secondary (backup)
process, which remains ready to take over if the primary process fails. During processing, the
primary process keeps the backup process informed of what it is doing (for example, sending a
request) by means of special interprocess messages, in an activity called checkpointing. Through
checkpointing, the backup process has enough information to take over and continue if the primary
process fails.
Both the PATHMON process and the TCP can be configured as process pairs to support Pathway
applications. When the PATHMON process is configured as a process pair, you are ensured the
ability to control and monitor OLTP system operation even if the primary PATHMON process fails.
When a TCP is configured as a process pair and the primary TCP fails, terminals controlled by the
TCP can still be used.
Pathway server-classes provide additional fault tolerance by allowing requests to be rerouted to
surviving server processes in a server-class if one of the server process fails.
Besides process pairs and server-classes, fault tolerance in a Pathway application is ensured by
the PATHMON process, the TCP, and the TMF subsystem. Using information stored in the PATHMON
configuration file, the PATHMON process automatically restarts processes at their initialization
level after a failure, allowing these processes to resume work immediately.
18 Introduction to Pathway Application Programming