Event Management in Distributed Control Systems

1
Event Management in Distributed Control Systems

• Gheorghe Sebestyen
• Technical University of Cluj-Napoca
• Computers Department

2
Content

• Introduction goals and motivations
• Event management in a distributed environment
• An Event management service model
• Analytical evaluation of the reaction time to
  events
• Conclusions

3
Events in control systems

• control systems
• event-driven reactive systems
• time-driven time (clock ticks) drives the
  system
• event handling
• in single computer systems
• Input/Output pooling
• Interrupts ISRs
• in distributed systems
• not a trivial task

4
Events in Distributed control

• problems concerning events
• detection, notification, ordering
• actions (tasks) triggering as response to events
• evaluation of time delays between event and
  action
• proposed solutions
• MMS (Manufacturing Message Specification)
• Virtual devices Event objects
• ROOM (Real-time Object Oriented Model)
• Interface objects for I/O and signaling
  operations

5
A system model based on Distributed services

• Distributed system a collection of autonomous
  and distributed services
• For a distributed control system
• data acquisition service
• resource management service
• supervision and control service
• time and synchronization service
• event management service

6
A fully distributed Event management service

7
The Event Management service

• Functions
• definition and configuration of data structures
  that describe and store events
• detection of conditions that trigger events
• dynamic event logging
• event notification mechanisms
• task scheduling based on the pending events-    
  event messaging
• event ordering and synchronization mechanisms
• remote access to event objects
• unique time reference for event time-stamping

8
Components of a Local Event Manager
9
Evaluation of the reaction-time to events

• The computational model
• tasks are statically allocated to network nodes
• every node has a task scheduler that use the Rate
  Monotonic scheduling algorithm this algorithm is
  optimal for a single computer system
• tasks communicate through messages, on the
  network
• messages are scheduled with a TDMA (Time Division
  Multiple Access) algorithm, which means that
  every node has a predefined time slice for
  communication
• a task or message is delayed by its predecessor
  (message or task) in the transaction

10
The reaction time for a task

• without communication
• ri Ci ? (?ri/Tj? )
• jgtprior
• where ri the reaction time of task i
• Ci the execution time of task i
• Tj the period of a task j that have a
  higher priority than task i
• if communication is involved
• ri Ji wi
• where Ji the launching jitter, caused by the
  predecessor message
• wi the reaction time caused by the task
  scheduling algorithm

11
The reaction time for a task (cont.)

• wi Ci ? (?(Jjwi)/Tj? Cj)
• jgtprior
• a recursive formula
• ri(k1) Ci ? (?ri(k)/Tj? Cj)
• jgtprior

12
Delivery time of a message

• rm ?(1 Im)/S? TTDMA
• where rm the delivery time of message m
• Im the number of messages that have higher
  priority than message m and which may be
  launched before message m
• S number of messages that can be transmitted
  by a node in its time slice of a TDMA period
• TTDMA the TDMA protocols period
• The term Im may be evaluated with the following
  expression
• Im ? ?(J j rm)/Tj?
• jgtprior
• where Jj the delay (jitter) caused by the
  task that generate the message
• Tj the period of a message j that have a
  higher priority than message m

13
Iterative computation

• R1 (m1) ?RM( J1(m))
• R2 (m1) ?RM( J2(m))
• ..
• Rn (m1) ?RM( Jn(m))
• Rret(m1) ?TDMA(Jret(m))
• J1(m1) ?1(Rret(m1))
• J2(m1) ?2(Rret(m1))
• ..
• Jn(m1) ?n(Rret(m1))
• Jret(m1) ?ret(R1(m1), R2(m1), Rn(m1))

14
Conclusions

• event management is a critical task in any
  control system
• distributed control systems require special event
  detection, notification and ordering mechanisms
• the service-based event management approach is
  proper for a distributed environment
• a scalable, reliable and fault-tolerance solution
• the reaction time of a distributed system can be
  evaluated with an analytical method