Temporal and Realtime Database: A Survey

1
Temporal and Real-time Database A Survey

• Gultkin Ozoyoglu
• Richard T. Snodgrass
• Speaker Keke Chen

2
Outline

• What is real-time database
• Time in real-time database
• Real-time datatransaction properties
• Implementation issues
• Conclusion

3
Whats real-time database?

• A real-time database is a database in which
  transactions have deadlines or timing
  constraints.
• Responding to queries
• Processing database updates
• Maintaining integrity

4
Time in Real-time Databases

• Valid time data items that are used to record
  time of external objects
• Transaction time setting value of a real world
  object is performed by a transaction.
• Time granularity

5
Real-time data and transaction properties

• Types of transactions in real-time db
• How to estimate transaction execution times?
• Factors that affect the transaction processing
• Consistencies in real-time database

6
Types of Transactions

• Hard transaction
• Missing a deadline is disastrous
• Soft transaction
• Negotiable deadline
• Firm transaction

7
Estimate worst case transaction execution times

• Values of data items they access
• Block or abort due to data and resource conflicts
• I/O scheduling or buffer management

8
Factors that affect transaction processing

• Deadline hard, soft or firm transaction
• Transaction arrival pattern Periodic, sporadic,
  aperiodic
• Data access type - random, predefined write only
  or read only
• Accessed object properties continuous or
  discrete
• Knowledge of items to be used
• CPU and I/O time

9
Consistencies in RTDB

• Internal consistency
• Data model
• Integrity constraints
• Serializable transaction executions
• External consistency
• Whenever a real-world object change its value,
  its counterpart data item also changes
• External consistency takes priority over internal
  consistency

10
Consistencies in RTDB (cont.)

• Transactions VS. Consistencies
• Cooperative, not competitive
• Serializabiliy must be extended or replaced
• Transactions trigger transactions
• Transaction scheduling in RTDB and process
  scheduling in OS

11
Implementation issues

• Transaction scheduling
• Concurrency control
• Buffer Management
• Disk I/O scheduling

12
Transaction scheduling

• With hard deadlines
• Scheduled with complete knowledge
• (arrival pattern, data access type, cpu, IO)
• With soft deadlines Priority assignment
  policies
• Earliest-deadline first, highest-value-first,
  least-slack-time first

13
Concurrency control

• Epsilon serializability
• Allow bounded inconsistency during conflict
  resolution
• Divergence control algorithm
• Conflicting access for read only transactions due
  to read-write conflicts
• High level concurrency
• Using application semantics

14

• Buffer Management
• Priority-LRU
• Priority DBMIN
• Disk I/O scheduling
• FD-SCAN (feasible deadline SCAN)
• Prioritized groups of disk requests
• And so on

15
Conclusion

• Future research areas
• Real-time data model
• Real-time transaction processing
• Real-time query language
• Real-time database design
• Performance evaluation