Page 197 - DCAP402_DCAO204_DATABASE MANAGEMENT SYSTEM_MANAGING DATABASE
P. 197
Database Management Systems/Managing Database
Notes forwards from the <checkpoint L> record till the end of the log. During the scan, perform redo
for each log record that belongs to a transaction on redo-list.
SQL does not have very specific commands for recovery but, it allows explicit COMMIT,
ROLLBACK and other related commands.
11.7 Buffer Management
When the database is updated, a lot of records are changed in the buffers allocated to the log
records, and database records. Although buffer management is the job of the operating system,
however, some times the DBMS prefer buffer management policies of their own.
Log Record Buffering
Log records are buffered in the main memory, instead of being output directly to a stable
storage media. Log records are output to a stable storage when a block of log records in the
buffer is full, or a log force operation is executed. Log force is performed to commit a transaction
by forcing all its log records (including the commit record) to stable storage. Several log records
can thus be output using a single output operation, reducing the I/O cost.
The rules below must be followed if log records are buffered:
1. Log records are output to stable storage in the order in which they are created.
2. Transaction T enters the commit state only when the log record <T commit> has been
i i
output to stable storage.
3. Before a block of data in the main memory is output to the database, all log records
pertaining to data in that block must be output to a stable storage.
These rules are also called the write-ahead logging scheme.
Database Buffering
The database maintains an in-memory buffer of data blocks, when a new block is needed, if the
buffer is full, an existing block needs to be removed from the buffer. If the block chosen for
removal has been updated, even then it must be output to the disk. However, as per write-ahead
logging scheme, a block with uncommitted updates is output to disk, log records with undo
information for the updates must be output to the log on a stable storage. No updates should be
in progress on a block when it is output to disk. This can be ensured as follows:
1. Before writing a data item, the transaction acquires exclusive lock on block containing the
data item.
2. Lock can be released once the write is completed. (Such locks held for short duration are
called latches).
3. Before a block is output to disk, the system acquires an exclusive latch on the block
(ensures no update can be in progress on the block).
A database buffer can be implemented either, in an area of real main-memory reserved for the
database, or in the virtual memory. Implementing buffer in reserved main-memory has
drawbacks. Memory is partitioned before-hand between database buffer and applications, thereby,
limiting flexibility. Although the operating system knows how memory should be divided at
any time, it cannot change the partitioning of memory.
190 LOVELY PROFESSIONAL UNIVERSITY
