Download presentation
Presentation is loading. Please wait.
Published byAlexander Fisher Modified over 8 years ago
1
© Virtual University of Pakistan Database Management System Lecture - 43
2
© Virtual University of Pakistan Write Sequence For a write operation entry is made in log file in RAM On commit, first, Database buffer is updated Log file is moved to disk log file Then write is performed from DB buffer is moved to disk
3
© Virtual University of Pakistan Recovery Sequence If system crashes On restart, recovery manager examines the log file Transactions with begin and commit to be redone No action for begin and aborted
4
© Virtual University of Pakistan Checkpoint Record Repetition, no harm How far in the log file should be redone Checkpoint record is inserted in log file at certain intervals
5
© Virtual University of Pakistan At Checkpoint Modified DB buffers to disk All log records from buff to disk Writing a checkpoint record to log; log record mentions all active transactions at the time
6
© Virtual University of Pakistan Recovery Routine Check last check point Any transaction committed before checkpoint, no action On or started after and commit: redone On and abort, or on and no end: no action
7
© Virtual University of Pakistan Summary of Deferred Updates Writes are deferred until commit for transaction is found For recovery purpose, log file is maintained Log file helps to redo the actions that may be lost due to system crash Log file also contains checkpoint recods
8
© Virtual University of Pakistan Immediate Updates Incremental log with immediate updates DB buffers are updated immediately, and files updated when convenient
9
© Virtual University of Pakistan Write Sequence Start Tr record in log, On write operation, write a log record Move log record to file Update the DB when convenient On commit write in log file
10
© Virtual University of Pakistan Recovery Sequence Trs with commit to redo, by copying new values Trs, with abort or no end undo, by copying the old value Repetition doesn’t matter Checkpoint record can be used
11
© Virtual University of Pakistan Summary of DB Recovery An improper shut down may damage database consistency Has to be detected and database to be brought in an consistent state
12
© Virtual University of Pakistan Log file contains records of all updates Updates may be deferred or immediate, record entries vary Transactions are redone or undone depending on situation Checkpoints help in efficient database recovery
13
© Virtual University of Pakistan Concurrency Control
14
© Virtual University of Pakistan Concurrent Access Sharing, one of the basic objective Multiple users accessing simultaneously is preferred Operations of more than one transactions can be performed simultaneously; interleaved transactions
15
© Virtual University of Pakistan Concurrent Access Two factors allow concurrent access Processors being very fast I/O controller can handle I/O Problems due to interleaving
16
© Virtual University of Pakistan Concurrency Control? Three problems due to interleaved transaction Lost Update Uncommitted Update Inconsistent Analysis
17
© Virtual University of Pakistan Lost Update Problem TIMETATBBAL t1t1 Read (BAL)1000 t2t2 …….Read (BAL)1000 t3t3 BAL = BAL - 50……..1000 t4t4 Write (BAL)950 t5t5 …….BAL = BAL + 10950 t6t6 ……..Write (BAL)1010
18
© Virtual University of Pakistan Database Management System Lecture - 43
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.