Issues in Building Real-Time Applications
郭大維 教授
ktw@csie.ntu.edu.tw
嵌入式系統暨無線網路實驗室
(Embedded Systems and Wireless Networking Laboratory)
國立臺灣大學資訊工程學系
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Introduction to Real-Time Systems
Checklist
⊕ What is a real-time system?
⊕ What is the way usually used to classify real-time tasks?
⊕ What are the issues and research for real-time systems?
⊕ Is there any misconception about real-time computing?
⊕ Is our current software development
environments suitable to time-critical systems?
⊕ What kinds of software architectures are adopted
or considered in current time-critical systems?
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Introduction to Real-Time Systems
What is a real-time system?Any system where a timely response by the computer to external stimuli is vital!
Examples: multimedia systems, virtual reality, games.
avionics, air traffic control, nuclear power plant
stock market, trading system, information access, etc.
What is a Real-Time System?
Does the definition make every computer a real-time computer?Yes! It is if we need some response from a computer within a finite time!!
Category of Real-Time Systems: Hard Real-Time Systems - catastrophic if some deadlines are missed.
Soft Real-Time Systems - otherwise.
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Issues in Real-time Computing
The field of real-time computing is especially rich in research problems!
However, real-time computing systems often differ from their counterparts in two ways:
More specific in their applications.
More drastic for their failures.
Real-Time Computing Computing
For example, CPU scheduling of tasks with different criticality!
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Structure of A Real-Time System - An Example
A control system
Rates -sensors & actuators, peripheral, control program
Phases -takeoff, cruise, and landing, etc.
sensors
actuators environment controlled
process
Task Execution
Clock
Display
operator
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Task Classes
Ways to classify real-time tasks: Predictability of their arrivals.
)Periodic tasks have regular arrival times.
)Aperiodic tasks have irregular arrival times.
• bounded inter-arrival time -> Sporadic tasks.
Criticality - consequences of non-timely executions.
)Critical tasks should have timely executions
• Most of them are hard real-time transactions
)Non-critical tasks are usually soft real-time tasks
• minimize miss ratio, minimize response time, maximize values contributing to the system, etc.
Issues and Research
Software engineering
System architecture, e.g., event-driven, time-line, time- driven, object-oriented, etc.
Network architecture, e.g., topology, predictability, and controllability.
Fault-tolerance and reliability evaluation, etc.
Tools for prototyping, simulation, code synthesis.
Operating systems
Task assignment and scheduling
Communication protocols
Failure management and recovery
Clock Synchronization, etc.
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Issues and Research
Programming languages
Better control over timing
Proper interface to special-purpose devices
Database systems
Concurrency Control
Failure recovery
Availability
Query Optimization, etc.
Specification and verification
Expressiveness and complexity
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Issues for Programming Environments
Loop size, timer granularity, imprecise timer, sleep(), multi-programming, etc.
Sequential programs, parallel programs, timely programs.
Client-server priority assignments - priority inversion.
Verification, analysis, and testing.Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Issues for Programming Environments
Potential Timing Hazards:Loop
…...
Sensor();
……..
computation……
……..
t = time();
SleepTime := ReadyTime + PERIOD - t;
ReadyTime = ReadyTime + PERIOD;
Sleep(SleepTime);
EndLoop;
Loop Size?
Time Elapsed Here?
Timer Granularity?
Real Sleep Time?
???Multiprogramming???
Issues for Programming Environments
The priority assignment for a Server TS? Processes TH and TL
Priority Inversion
TS
TL
TH
TS
TL
TH
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Issues for Timer Drifting
Your timer settings
Timer interrupts
Starts of periods
0 1 2 3 4 5 6
0 1 2 3 4,5 6
Set the timer resolution to x.
Round off all of timeout intervals to integer multiples of x!
Software Architectures and Fault Tolerance Issues for Real-Time Applications
Source: C. Douglass Locke & Farnam Jahanian, RTCSA’96 Talks Presentation.
郭大維 教授
ktw@csie.ntu.edu.tw
嵌入式系統暨無線網路實驗室
(Embedded Systems and Wireless Networking Laboratory)
國立臺灣大學資訊工程學系
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Software Architectures for Real-Time Applications
Popular architectures: Timeline (i.e., cyclic executive or frame-based)
Event-driven (with both periodic and aperiodic activities)
Pipelined
Client-Server
Impacts Performance and life-cycle cost
Critical design decisions such as synchronization and exceptions.
No restriction on parallel processing.Timeline or Cyclic Executive
A major cycle consists of a non-repeating set of minor cycles Operations are implemented as procedures.
The timer calls each procedure in the list.
No concurrency exists.
Very high life-cycle cost but very predictable in the run-time behavior!Timeline or Cyclic Executive
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Event-Driven
Characteristics: Trigger schedulable tasks by I/O completion and timer events.
Task Priority: Determined by timing constraints, e.g., RMS, or by semantic importance.
Ways to avoid synchronization is needed for predictable response.
Processor utilization is preserved foraperiodic events for response predictability.
Prone to event shower! Good for systems with
spare computation power!
Event-Driven
Persistent data shared within partitions
Pipelined
Characteristics: Trigger schedulable tasks by I/O completion, timer events, and inter-task messages.
The system can be described as a set of pipelines of task invocations.
Task priority Increasing task priorities in a unidirectional
pipeline will minimize the message queue buildup.
Equal task priority setup is normal for bi- directional pipelines.
Prone to event shower! Good for systems with spare computation power!Pipelined
No shared data -- persistent data in Abstract Data Objects
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Client-Server
Characteristics: Trigger schedulable tasks by I/O completion, timer events, and inter-task messages.
Control flow for an event stays at a node while data flow is distributed.
Task priority Priority inheritance is used ideally. Practically task priorities are set equally, and message priorities are used instead to avoid bottlenecks.
More messages are exchanged but are
significantly easier in debugging than pipelined
systems.
Client-Sever
No shared data--persistent data in Abstract Data Objects
Fault Tolerance
Definition: A real-time fault-tolerance system is a system that can deliver its service even in the presence of faults.
Timeliness versus Fault Tolerance Possible Faults: Hardware/software errors,
violation of timing constraints because of
the “environment”.
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
Fault Tolerance
Use redundancy to detect errors and mask failures Space Redundancy: replication of physical devices.
Time Redundancy: repetition of a computation or communication.
Information Redundancy: specific encoding scheme, e.g., parity bit.
Copyright: All rights reserved, Prof. Tei-Wei Kuo, Embedded System and Wireless Networking Lab, National Taiwan University.
