Concurrency is a concern to all database administrators and developers. Even if you are working with a single-user database, you must be concerned with concurrency to process requests from multiple applications or even from multiple connections from a single application. These applications and connections can interfere with each other in exactly the same way as multiple users in a network setting.

The way you group SQL statements into transactions can have significant effects on data integrity and on system performance. If you make a transaction too short and it does not contain an entire logical unit of work, then inconsistencies can be introduced into the database. If you write a transaction that is too long and contains several unrelated actions, then there is a greater chance that a ROLLBACK could unnecessarily undo work that could have been committed quite safely into the database.

If your transactions are long, they can lower concurrency by preventing other transactions from being processed concurrently.

There are many factors that determine the appropriate length of a transaction, depending on the type of application and the environment.

• SQL Anywhere database servers