p. 256

1-3-3 Database Control Function

Database control function is a function for improving reliability and security of the data recorded in the databases.

(1) Maintenance function

Maintenance function is a function for maintaining integrity of data (i.e., data integrity). The following double update is the typical example where integrity of data is lost. (As a result, inconsistency occurs.)

Exclusive control is designed for resolving such a problem. There are several methods in exclusive control, and the typical method is lock system. The lock system locks the database record that is referenced and restricts further referencing until the process is completed. When this method is used, there is no risk of data being referenced by another program during the update, therefore, the integrity of data increases. The following two types of locks are used separately in the lock system. · 1 question

p. 257

However, the problem called deadlock may occur in the lock system. Deadlock is the phenomenon where multiple programs are simultaneously in the waiting state because of the lock and their executions completely stop. It is difficult to fully avoid the deadlock. But, by reducing the unit (granularity) of applying the lock, the possibility of the occurrence of deadlock can be reduced. · 2 questions

p. 258

Other methods of exclusive control include semaphore system, which manages databases as resources by means of semaphore.

(2) Security protection function

Security protection function is a function for maintaining security of data (i.e., data security). Critical data and highly confidential data are stored in databases. Therefore, data security protection becomes necessary. The following security protection functions are provided by DBMS.

p. 259

(3) Failure recovery function

Failure recovery function is a function that restores databases at the occurrence of a failure. Failure recovery uses files, such as backup file where a database at a particular time is copied as it is, and journal file (i.e., log file) where an update process performed on the database is recorded. Journal records are created in the memory and are written in a file at the timing of commit (i.e., process of finalizing the update process) which is performed when the database update process for each transaction is completed. A log is prepared by setting a checkpoint for a certain time interval (or a certain processing amount). The database being updated in the memory is recorded in the checkpoint file. With this, it is possible to recover from the checkpoint time when a failure has occurred. · 2 questions

p. 260

The following are two types of recovery processes performed for failure recovery.

The recovery process is also performed for each transaction. In this case, transactions committed after a checkpoint are recovered to the status after completion of commit with rollforward. On the other hand, for the transactions that are running at the time the failure occurred, processes that are not committed are canceled (i.e., rolled back) and transactions are executed again after the database is reverted to the status before the process started. In the case of Figure 4-8, T1 and T2 committed after the checkpoint are recovered with rollforward from the checkpoint time. Meanwhile, T3, which was being updated at the time the

p. 261

failure occurred, is executed again when the database is reverted to the status before the process started with rollback from the checkpoint time.

The following two methods are available for restarting the database.

A repetitive update (i.e., addition or deletion) of the database may result in a wasteful storage area that is not reused, which may decrease the access efficiency. In such a case, reorganization that optimizes the database is also one of the failure recovery functions.