Database Chapter 20: Transaction Processing

Introduction to Transaction Processing Concepts

• A transaction represents a single unit of database processing.
• Transaction processing systems are designed to handle large databases with numerous concurrent users, emphasizing high availability and rapid response times.

Types of Database Management Systems

• Single-user DBMS:
  • Limited to one user at a time.
  • Example: Home computers.
• Multiuser DBMS:
  • Allows multiple users to access the system simultaneously.
  • Example: Airline reservation systems, utilizing multiprogramming concepts.

Multiprogramming and Processing

• Multiprogramming enables concurrent execution of multiple processes through the operating system.
• Interleaved processing refers to handling multiple transactions by alternating between them.
• Parallel processing involves executing multiple transactions simultaneously.

Transaction Components

• A transaction consists of begin and end statements to define its boundaries.
• Types of transactions include:
  • Read-only transactions: Do not modify data.
  • Read-write transactions: Both read and modify data.

Database Items and Granularity

• A database consists of named data items, with size varying based on its granularity (record, disk block, or attribute value).
• Concepts of transaction processing are irrelevant to the granularity of data items.

Read and Write Operations

• read_item(X): Reads the specified database item and stores it in a program variable.
• write_item(X): Writes the value from a program variable to the database item.
• Operations involve locating, copying, and possibly updating disk blocks in memory.

DBMS Buffers

• DBMS maintains multiple memory data buffers for efficient data access.
• Buffer replacement policies (e.g., least recently used) are employed when memory is full.

Concurrency Control Issues

• Transactions may run concurrently, potentially leading to data integrity issues such as:
  • Lost Update Problem: Concurrent access results in some updates being lost.
  • Temporary Update Problem: Involves reading uncommitted changes.
  • Incorrect Summary Problem: Results from aggregate data being affected by concurrent transactions.
  • Unrepeatable Read Problem: Changes made by another transaction disrupt read consistency.

Importance of Recovery

• Transactions need recovery mechanisms to handle failures:
  • Committed transactions: Changes persist permanently.
  • Aborted transactions: Changes are discarded without impact.
• DBMS must ensure all operations of a transaction are either completely applied or not applied at all.

Common Types of Transaction Failures

• Computer Failures: Due to hardware or software crashes.
• System Errors: Including logic errors and user interruptions.
• Local Errors: Such as data access failures.
• Concurrency Violations: Issues like deadlocks and serializability violations.

Less Common Transaction Failures

• Disk Failures: Issues related to disk read/write mechanisms.
• Catastrophic Failures: Severe power or environmental disruptions.

Transaction State Management

• The system monitors each transaction's lifecycle:
  • Identifying when transactions begin, read/write data, and commit/abort.

System Logging

• The system log tracks transaction operations in a sequential manner, unaffected by most failures.
• A log buffer in memory stores transactions until full, then transfers them to the disk log.
• Periodic backups of the log file ensure data integrity and recovery options through undo and redo capabilities.

Sample Log File Entries

• Example entries include:
  • [start_transaction, T]: Indicates start of transaction T.
  • [write_item, T, X, old_value, new_value]: Indicates a change from old to new value for database item X.
  • [read_item, T, X]: Indicates transaction T has read item X.
  • [commit, T]: Indicates successful completion of transaction T.
  • [abort, T]: Indicates transaction T was aborted.