CONCURRENCY CONTROL

Problem Definition
1. Objective
  1. Maximize transaction throughput(work perform) while preventing interference among multiple users
  2. Throughput: number of transactions processed per unit time eg >1million transaction per second
2. Constraint
  1. No interference: result same as serial/sequential effect
  2. Interference occurs on commonly manipulated data known as hot spots
Concurrency Control Problem
1. Lost Update Problem e.g:slide no.11&12
  1. Preventing Lost Update Problem Using 2PL e.g:slide no.28
2. Uncommitted Dependency Problem e.g:slide no.13&14
  1. Preventing Uncommitted Using 2PL e.g:slide no.29
Concurrency Control Tools
1. Locking
  1. Obtain lock before accessing an item
  2. Wait if a conflicting lock is held
    1. Shared lock (S lock): conflicts with exclusive locks
    2. Exclusive lock (X lock): conflicts with all other kinds of locks
  3. Locking Conflict e.g:slide no.22
  4. Locking Granularity e.g:slide no 23
2. Deadlock
  1. Deadlock(Mutual Waiting) e.g:slide no.24
  2. Deadlock Resolution
    1. Detection
      1. Overhead is reasonable for deadlocks among 2 or 3 transactions
    2. Timeout
      1. Waiting limit
      2. Can abort transactions that are not deadlocked
      3. Timeout interval is difficult to determine

Zusammenfassung der Ressource

	Erstellt von azzaiqmal vor fast 11 Jahre