Data and Database Administration Chapter 12 Contd n

Data and Database Administration Chapter 12 (Contd. )

n Recovery and restart procedures

Transaction Integrity n n Many DB transactions result in changes to one or more DB records. If an error occurs during a transaction processing, then some form of DB recovery will be required. To understand DB recovery, we must understand the concept of transaction integrity. A business transaction represents a business activity which has a sequence of steps. When processing transactions, DBMS must ensure the four basic properties called ACID: q Atomic n meaning that the transaction can not be subdivided, it may be processed in entirety or not at all.

Transaction Integrity (Contd. ) q Consistent n q Isolated n q meaning that DB constraints that must be true before the transactions must also be true after the transaction. meaning changes to the DB are not revealed to the users until the transaction is committed. Durable n meaning that changes are permanent, once a transaction is completed, no subsequent failures can revert back the effect of transaction.

Recovery and Restart Procedures n Type of recovery procedure depends on q q q n The nature of failure The sophistication of DBMS recovery facilities Operational policies and procedures Techniques frequently used are: q q Switch Restore/run Backward recovery Forward recovery

Recovery and Restart Procedures: Switch n n To switch to an existing copy of the database, the database must be mirrored, i. e. two copies of the database must be kept and updated simultaneously. On failure, processing is switched to the duplicate copy of the database. This technique allows fastest recovery. Level-1 RAID systems implement mirroring. q q Instead of storing the DB on a large drive, it is distributed across several smaller, less expensive disks and mirrored to a second set of such disks. On disk failure, system switches to the mirrored disks.

Recovery and Restart Procedures: Restore/run n This technique involves reprocessing the day’s transactions (upto point of failure) against the backup copy of the database/portion of DB. q q q n Advantage q n The database is shut down then most recent copy of DB to be recovered is mounted, and all transactions that have occurred since that copy are rerun. simplicity (no need to create DB change journal, no restart procedures) Disadvantages q q reprocessing overhead (depending on frequency of making backup copies, several hours may be required) Processing of new transactions will have to be deferred until recovery is completed. The sequence in which transactions are rerun, may lead to quite different results. Hence, restore/rerun is used only as a last resort in DB processing.

Backward Recovery (rollback) n n n Through this technique, the DBMS back outs or undoes the unwanted changes to the database. Before-images of the records are applied to the database. Thus, the DB returns to its earlier state. Backward recovery is used to reverse changes due to aborted or abnormally terminated transactions.

Forward Recovery (rollforward) n n n A technique that starts with an earlier copy of the database. After-images (results of good transactions) are applied to the database, and DB is quickly moved forward to a later state. It is much faster & accurate as compared to restore/rerun because: q q It does not involve the time consuming reprocessing of transactions Only the most recent after images need to applied. A DBrecord may have a series of after-images but the most recent, good after-image is used in rollforward.

Types of DB failure n Four of the most common types of failures/problems are: q q Aborted transactions Incorrect (but valid) data System failure Database loss or destruction