1. Distributed Real-Time systems 1 By: Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Computer Networks

2. Outline  Real-time system  Time Properties  Example for Real-time  Types of Real-time  Distributed Real-Time Systems  What is distributed system  Resource management  Global scheduling  Transfer policy  Selection policy  Location policy  Information policy 2  Scheduling Concepts  Objective  Properties  Performance measures  A simple model  Scheduling Algorithms

3. 3 Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks Real-time system  real-time systems are defined as those systems in which the correctness of the system depends not only on the logical result of computation, but also on the time at which the results are produced.  A real-time system will usually have to meet many demands within a limited time.  a real-time system consists of a controlling system (computer) and a controlled system (environment).  The controlling system interacts with its environment based on information available about the environment.

4. 4 Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  a real-time system consists of a controlling system (computer) and a controlled system (environment).  The controlling system interacts with its environment based on information available about the environment. Real-time system

5. 5 Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Release time (or ready time): Time at which the task is ready for processing.  Deadline: Time by which execution of the task should be completed, after the task is released.  Minimum delay: Minimum amount of time that must elapse before the execution of the task is started, after the task is released.  Maximum delay: Maximum permitted amount of time that elapses before the execution of the task is started, after the task is released.  Worst case execution time: Maximum time taken to complete the task.  Run time: Time taken without interruption to complete the task, after the task is released.  Weight (or priority): Relative urgency of the task. Time Properties

6. 6 Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks Example For Real-Time  Real-time systems span a broad spectrum of complexity from very simple micro-controllers to highly sophisticated, complex and distributed systems.  Some examples of real-time systems include :  process control systems,  flight control systems,  flexible manufacturing applications,  robotics,  intelligent highway systems,  and high speed  and multimedia communication systems  Anti-Brake System (ABS)

7. 7 Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks Types Of Real-Time  soft real-time tasks  firm real-time tasks  hard real-time tasks

8. 8 Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Objective  Properties  Performance measures  A simple model  Scheduling Algorithms  Uniprocessor Scheduling Algorithms  Multiprocessor Scheduling Algorithms  Distributed Real-Time Systems Scheduling Consepts

9. 9 Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Meeting the timing constraints of the system  Preventing simultaneous access to shared resources and devices  Attaining a high degree of utilization while satisfying the timing constraints of the system  Reducing the cost of context switches caused by preemption  Reducing the communication cost in real-time distributed systems; we should find the optimal way to decompose the real-time application into smaller ortions in order to have the minimum communication cost between mutual portions  Covering reliability, security, and safety. Objective

10. Properties Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Soft/Hard/Firm real-time tasks  Periodic/Aperiodic/Sporadic tasks  Preemptive/Non-preemptive tasks  Multiprocessor/Single processor systems  Fixed/Dynamic priority tasks  Flexible/Static systems  Independent/Dependent tasks

11. performance measures Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Depending on the type of application we are confronted with, different performance measures or optimality criteria are used to evaluate schedules.  Among the most common measures in scheduling theory are :  schedule length   mean flow time   mean weighted flow time

12. A Simple Model Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Let the deadline Di for each task Ti be Pi TaskPriorityPeriod Computation time T1172 T22164 T33317

13. Scheduling for the Simple Model Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Priorities with preemption

14. Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Priorities without preemption Scheduling for the Simple Model

15. 15 Scheduling Algorithms of Real-Time Systems EDF LLF RM DASA LBESA

16. Distributed Real-Time Systems Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  What is distributed system  Resource management  Global scheduling  Transfer policy  Selection policy  Location policy  Information policy

17. What is distributed system? Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  A set of nodes commun. through a network  Network could be LAN or WAN  Nodes could be homogeneous or heterogeneous N1 Network (WAN/LAN) N2 N3 Nn

18. Why distributed systems? Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Applications themselves are distributed o E.g., command and control, air traffic control  High performance o Better load balancing  High availability (fault-tolerance) o No single point of failure

19. problems with distributed systems Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Resource management is difficult o No global knowledge on workload o No global knowledge on resource allocation  No synchronized clock (or clocks need to be synchronized)  Asynchronous nature of the nodes  Communication related errors o Out of order delivery of packets, packet loss, etc.

20. System model Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  The application is realized on a distributed system  Tasks arrive at each node independent of other nodes  Each node has resource manager for managing the workload at local node and for facilitating migration of workload to remote nodes  Nodes cooperate among themselves for meeting tasks’ deadlines

21. Workload assumptions Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Periodic tasks and aperiodic tasks  Periodic messages and aperiodic messages  Task may have precedence constraints  The commn. pattern among two communicating periodic tasks is also periodic  Two communicating tasks could be scheduled on two different nodes  Meeting tasks deadlines require bounding and meeting message deadlines

22. Resource management in Distributed RT systems (Node architecture) Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Local scheduling  Resource management within a node  Task scheduling, resource reclaiming, etc.  Global scheduling  Balancing load across nodes  Transfer policy, selection policy, information policy, and location policy  Communication resource management  QoS routing (channel setup time)  Resource reservation (channel setup time)  Packet scheduling (run-time)

23. Global scheduling Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Goal: migrate tasks from a local node (when it is heavily loaded) to a lightly loaded node  Transfer policy: when tasks are to be migrated from/to local node to/from remote nodes  Selection policy: which tasks are to be migrated  Location policy: where tasks are to be migrated  Information policy: what information is exchanged among nodes to realize task migration

24. Transfer policy Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Load index: the quantitative measure of node’s load  Non-real-time systems: queue length, processor utilization  Real-time systems: processor utilization, tasks’ laxity/deadline  Transfer policy determines whether the current node is suitable to participate in a task migration either as a sender or as a receiver  Threshold-based load index  Two thresholds (L-upper and L-lower) based on which a node’s load is classified as Light, Normal, or Overload  Light load implies the node could be a receiver for task migration  Heavy load implies the node is a sender for task migration  Normal load implies neither sender nor receiver  Fixing thresholds is hard

25. Selection policy Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Once transfer policy determines the current node is the sender of a task migration, selection policy decides which tasks to migrate  While choosing the tasks, following needs to be considered  End-to-end delay: sum of local decision time, migration time, remote decision time, and task’s execution time must be less than task’s deadline  Task’s affinity to node – e.g., the required resource must be available at the remote node  Task’s “value” – it is better meet deadlines of higher value offering tasks

26. Location policy Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Choosing the receiver node for a task migration  There are several policies possible  Random policy – select the receiver randomly  Polling policy – poll the potential receivers of their load in sequential or parallel  Information based – based on the information provided by the information policy

27. Information policy Mahdi Sadeghizadeh Website: Sadeghizadeh.ir Advanced Networks  Nodes exchange state info so as to obtain global state  Demand-driven policy  A node collects state info from other nodes when it becomes a sender or receiver for task migration  Depends on node’s load state change to Light or Heavy  State-driven policy  Whenever node’s load state changes, it informs other nodes  Similar to other demand-driven  Periodic policy  Nodes periodically exchange state info irrespective of their states

28. Thank You 28