1 / 9

Recovery Concepts

Recovery Concepts. Chapter 18 (lightly). Occasions for Recovery. Non-catastrophic: need Log only Transaction abort "Normal" system crash Catastrophic: need Log + Backup Physical loss of disks Pervasive application error

tender
Download Presentation

Recovery Concepts

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Recovery Concepts Chapter 18 (lightly)

  2. Occasions for Recovery • Non-catastrophic: need Log only • Transaction abort • "Normal" system crash • Catastrophic: need Log + Backup • Physical loss of disks • Pervasive application error • System software error (corrupted filesystem, buffer management errors, etc.) • Virus, sabotage, etc.

  3. ARIES • A particular recovery algorithm emphasized in the book • Actually it uses many standard techniques • Strategies and concepts more important that ARIES per se

  4. (Review) The Log File • Needs to be on "stable storage" • Contains: • Transaction starts/stops • DB writes: "before" and "after" images • befores can be used to rollback an aborted transaction • afters can be used to redo a transaction without reexecuting it • COMMITs and ABORTs The log itself is as critical as the DB! Reliable backups are critical, too!

  5. Strategies Which Anticipate Normal Recovery • Deferred update • Writes are not actually applied to DB until after T commits. • No UNDO is needed. • Implementation: buffers, shadow page table, etc. • Immediate update • Writes are actually applied as T executes • Aborted transactions: UNDO (rollback)

  6. System Crash (without loss of DB) • Assume that ALL uncommitted T's have aborted • Procede as for transaction abort • Allow no new transactions until recovery is complete

  7. Catastrophe • First: restore from a full backup • Then: Roll forward from log • REDO all committed transactions • Apply all logged WRITEs • Could actually REDO changes in reverse chrono order: i.e., only apply latest change • T's interrupted by the catastrophe must be restarted or user notified

  8. Checkpoints • Checkpoint: "snapshot" of a system at a particular time • Might include: info about which T's are running; which pages are "dirty", etc. • Use: May be able to recover from the last checkpoint (rather than reprocessing entire log)

  9. Disaster Recovery via Redundancy • A reliable duplicate copy could be used for "instant" recovery • copy could be "hot" (in use by applications) or only on standby • SW-based • managed by DBMS or OS • could be part of a distributed system • HW-based • RAID: Redundant Array of Inexpensive Disks

More Related