Presented By: Jeff Reyen

1
Safety Reliability Design Patters for
Distributed/Real Time Systems

• Presented By Jeff Reyen
• Kshirsagar Naidu
• Supun Wijenayake
• Anika Choudhury

2
Safety and Reliability

• Safety and reliability have one important aspect
  in common their handling requires redundancy
• Some Safety and Reliability Design Patterns
• Triple Modular Redundancy
• Monitor-Actuator Pattern
• Watch Dog Pattern
• Safety Executive Pattern

3
Triple Modular Redundancy

• Protection against random fault with continuation
  of functionality
• Two types of faults Random and Systematic
• TMR is a pattern used to enhance reliability and
  safety in situations where there is no fail safe
  state
• A fail safe state is a state which is known to be
  safe usually off or depowered.

4
Triple Modular Redundancy
INPUT
OUTPUT
5
Pros Cons Of TMR

• Cons
• TMR can only detect random faults
• Comparer may fail causing the whole system to
  fail
• High Cost
• Pros
• TMR is time efficient in presence of faults

6
TMR

• The TMR pattern is a common in applications where
  reliability needs are quite high.
• Therefore worth the cost of Replicating hardware
  and software

7
Monitor-Actuator Pattern

• A special case of Diverse Redundancy Pattern.
• Consists of a Monitor and Actuator channel.
• Monitor channel used to check for
  inconsistencies.
• Safety at low cost option.
• Requires a fail safe state.

8
Monitor-Actuator Pattern

• System halts if problem in Actuator channel.
• System can operate while problem in Monitor
  channellatent fault.
• The Monitor channel needs to be periodically
  checked.
• life tics sent between channels to indicate
  health of other system.

9
Monitor-Actuator Structure Diagram
10
Monitor-Actuator Sequence Diagram
11
Monitor Actuator Pros/Cons

• Pros
• relatively inexpensive safety solution
• can operate against both systematic and random
  faults
• Cons
• can only be used for systems with fail-safe
  states
• not suitable for systems with high availability
  requirements
• system cannot function after a fault is
  identified

12
Watchdog Pattern

• Purpose
• To check if processes are executed in timely
  manner.
• Watchdog is lightweight and inexpensive
• For many real-time systems, watchdog is combined
  with heavy weight safety/reliability pattern.

13
Watchdog Pattern

• Problem
• RT systems timely events have hard deadline
• If an event occurs after its expected, the event
  is erroneous
• Eg PID controllers are sensitive to time lag
  between input signal and output.

14
Watchdog Pattern Component Diagram
15
Watchdog Pattern

• How it works
• Waits for Actuation Channel to send and expected
  signal (called a stroke) in a timeframe.
• Most watchdogs dont concern when the stroke
  comes too quickly
• If valid stroke gt invoke integrity test.
• If invalid stroke gt shutdown or restart

16
Watchdog Pattern Summary

• Lightweight inexpensive
• Purpose is to check if processes are computed in
  time based manner, therefore pattern doesnt
  concern in correctness of process.
• Watchdog is not integrated with main system.
• Not suitable for safety critical systems, needs
  to be combined with heavier design patterns.

17
Safety Executive Pattern

• A sophisticated extension of Watchdog pattern
• Tracks and coordinates all safety monitoring
  actions.
• Takes as inputs
• - Watchdog time-outs
• - software error assertions
• - faults identified by Monitor-Actuator Pattern
• - continuous or periodic Built-in-tests (BIT)

18
Safety Executive Pattern
19
Safety Executive Pattern
20
Safety Executive Pattern

• Pro
• -Isolates the higher level application
  from the safety issues and prevent
  extra checks and recovery actions.
• Con
• - complex design
• -expensive to implement

21
Summary

• Importance of safety and reliability of a system.
• Design Pattern
• -problem context
• -solution
• -consequences
• TMA -gt protection against random faults with
  continuation of functionality.
• M-A -gt protection against random and system
  faults with a fail-safe state.
• Watchdog -gt lightweight version of M-A pattern.
• SE -gt safety of complex systems with complex
  mechanisms to achieve fail-safe
  sates.

22
Thank you!

• Questions/Comments??

23
References

• Real-Time Design Patterns robust scalable
  architecture for Real-time systems / Bruce Powel
  Douglass.
• http//paula.oulu.fi/shickey/RTSD/RTSD_Lect7_2007
  .pdf
• http//soclab.ee.ntu.edu.tw/socpc/course_slides/2A
  rchitecturalDesign.pdf
• http//www-md.e-technik.uni-rostock.de/ma/gol/ilog
  ix/scritd.pdf
• http//www.uml.org.cn/sjms/pdf/ThesisFinal.pdf
• http//www.empress-itea.org/deliverables/D2.1-2.2_
  Appendix_B_v1.0_Public_Version.pdf