Misconceptions About Real

Download Report

Transcript Misconceptions About Real

Misconceptions About RealTime Databases
IEEE Computer
Authors: John Stankovic, Sang Hyuk Son,
Jorgen Hansson
Presented By: Patti Kraker
Definitions

Real-time database is a database
system





Queries
Schemas
Transactions
Concurrency control support
Storage management
Definitions

Timing constraints associated with
transactions


Not every transaction, but some
Data valid for specific time intervals
Definitions

Time semantics on data



E.g. sensor data, stock market prices
Semantics indicate valid for certain time
period
In a real-time database

Validity interval part of database schema
Definitions

Transaction Correctness

Absolute time consistency


Relative time consistency


Individual data is temporally valid to acceptable
degree of accuracy
Multiple data items updated within specified
time interval of each other
E.g. time and pressure
Misconceptions


Real-time systems synonymous with
speed
Current database technology can be
used in real-time database systems
Misconceptions/Research


Real-time database properties:
temporality, predictability, specialization
Research areas for real-time databases
Real-time systems
synonymous with speed
Hardware advances

Hardware advances will address realtime database requirements

Parallel processing



Does not mean meet timing constraints
Inc. size and complexity
Hardware alone not ensure data accurate
Hardware advances

Sum up hardware advances

Transaction that uses obsolete data more
quickly is still incorrect
Advanced DB technology

Advancing Standard Protocols

Better buffering, faster commit protocols,
better query processing


Can not guarantee deadlines or valid data
Required advances include:


time cognizant protocols, commit processing
transaction scheduling, logging, recovery
Advanced DB technology

Sum up advanced db technology

These protocols better at supporting realtime transaction and data correctness than
standard database protocols that simply go
fast.
Real-time = Fast


Fast computing: Aim to minimize
response time of transactions
Real-time computing: Aim to meet
timing constraints and data validity of
transactions and keep database current

Need time cognizant protocols
Current DB technology can
solve real-time problems
Traditional databases

Traditional databases can handle realtime

Can define field that contains validity
interval


Every transaction must check this field to
ensure absolute and relative validity
Can run earliest deadline scheduling
through priority of transaction
Traditional databases


By adding those features, moving
toward real-time db
More efficient to build them into system
than force fit traditional db
Db in main memory

Placing a conventional database in main
memory is sufficient (non-real-time db designers)


Does eliminate disk delays
Still unpredictable:



Delays due to blocking on locks
Transaction scheduling
Stolen processing time to handle external
interrupts
Db in main memory

Sum up current db technology

Increases in performance can not
completely make up for lack of timecognizant protocols
Real-time DB in main memory

A real-time db must reside totally in
main memory (real-time db designers)


Reasons: to increase speed, avoid seek
and rotational delays of disks
Primary issue I/O


Minimize response time, maximize throughput,
maintain fairness
Typical disk scheduling algorithms are FirstCome-First-Served (FCFS), Shortest-Seek-TimeFirst (SSTF), etc.
Real-time DB in main memory

Sum up real-time db in main memory

Since deadline and importance of
transaction are not considered when disk
request are scheduled, timeliness of
transaction is jeopardized.
Start of real-time databases

Real-time databases emerged from:

Telecommunications, manufacturing, and
avionics applications

Conventional databases not adequate
Research of real-time db


Increased audio and video applications
More research problems being
addressed