1. Silberschatz, Galvin, and Gagne  1999 6.1 Applied Operating System Concepts Module 6: CPU Scheduling 10/19/03 Basic Concepts Scheduling Criteria Scheduling Algorithms Multiple-Processor Scheduling Real-Time Scheduling Algorithm Evaluation NOTE: Instructor annotations in BLUE

2. Silberschatz, Galvin, and Gagne  1999 6.2 Applied Operating System Concepts Basic Concepts Objective: Maximum CPU utilization obtained with multiprogramming A process is executed util it must wait - typically for completion if I/O request, or a time quanta expires. CPU–I/O Burst Cycle – Process execution consists of a cycle of CPU execution and I/O wait. CPU burst distribution

3. Silberschatz, Galvin, and Gagne  1999 6.3 Applied Operating System Concepts Alternating Sequence of CPU And I/O Bursts

4. Silberschatz, Galvin, and Gagne  1999 6.4 Applied Operating System Concepts Histogram of CPU-burst Times “Tune” the scheduler to these statistics

5. Silberschatz, Galvin, and Gagne  1999 6.5 Applied Operating System Concepts CPU Scheduler Selects from among the processes in memory that are ready to execute, and allocates the CPU to one of them. CPU scheduling decisions may take place when a process: 1. Runs until it Switches from running to waiting state … stop executing only when a needed resource or service is currently unavailable. 2.Switches from running to ready state in the middle of a burst – can stop execution at any time 3.Switches from waiting to ready. … ? 4.Runs until it Terminates. Scheduling under 1 and 4 is nonpreemptive. All other scheduling is preemptive. Under nonpremptive scheduling, once a CPU is assigned to a process, the process keeps the CPU until it releases the CPU either by terminating or switching to wait state - “naturally” stop execution.

6. Silberschatz, Galvin, and Gagne  1999 6.6 Applied Operating System Concepts Dispatcher Dispatcher module gives control of the CPU to the process selected by the short-term scheduler; this involves: –switching context –switching to user mode –jumping to the proper location in the user program to restart that program Dispatch latency – time it takes for the dispatcher to stop one process and start another running. Both scheduler and dispatcher are performance “bottlenecks” in the OS, and must be made as fast and efficient as possible.

7. Silberschatz, Galvin, and Gagne  1999 6.7 Applied Operating System Concepts Scheduling Criteria CPU utilization – keep the CPU as busy as possible Throughput – # of processes that complete their execution per time unit Turnaround time – amount of time to execute a particular process Or: the time from time of submission to time of completion – includes waits in queues in addition to execute time. Waiting time – amount of time a process has been waiting in the ready queue – sum of the times in ready queue - this is from the point of view scheduler - scheduler does not look at CPU time or I/O wait time (only ready queue time) if it minimizes “waiting time”... Get a process through the ready queue as soon as possible. Response time – amount of time it takes from when a request was submitted until the first response is produced, not output (for time-sharing environment)

8. Silberschatz, Galvin, and Gagne  1999 6.8 Applied Operating System Concepts Optimization Criteria Max CPU utilization Max throughput Min turnaround time Min waiting time Min response time

9. Silberschatz, Galvin, and Gagne  1999 6.9 Applied Operating System Concepts First-Come, First-Served (FCFS) Scheduling Example:ProcessBurst Time P 1 24 P 2 3 P 3 3 Suppose that the processes arrive in the order: P 1, P 2, P 3 The Gantt Chart for the schedule is: Waiting time for P 1 = 0; P 2 = 24; P 3 = 27 Average waiting time: (0 + 24 + 27)/3 = 17 P1P1 P2P2 P3P3 2427300

10. Silberschatz, Galvin, and Gagne  1999 6.10 Applied Operating System Concepts FCFS Scheduling (Cont.) Suppose that the processes arrive in the order P 2, P 3, P 1. The Gantt chart for the schedule is: Waiting time for P 1 = 6; P 2 = 0 ; P 3 = 3 Average waiting time: (6 + 0 + 3)/3 = 3 Much better than previous case. Convoy effect short process behind long process P1P1 P3P3 P2P2 63300

11. Silberschatz, Galvin, and Gagne  1999 6.11 Applied Operating System Concepts Shortest-Job-First (SJF) Scheduling Associate with each process the length of its next CPU burst. Use these lengths to schedule the process with the shortest time. Two schemes: –nonpreemptive – once CPU given to the process it cannot be preempted until completes its CPU burst. –Preemptive – if a new process arrives with CPU burst length less than remaining time of current executing process, preempt. This scheme is know as the Shortest-Remaining-Time-First (SRTF). SJF is optimal – gives minimum average waiting time for a given set of processes.

12. Silberschatz, Galvin, and Gagne  1999 6.12 Applied Operating System Concepts ProcessArrival TimeBurst Time P 1 0.07 P 2 2.04 P 3 4.01 P 4 5.04 SJF (non-preemptive) FCFS is “tie breaker” if burst times the same. Average waiting time = (0 + 6 + 3 + 7)/4 - 4 Example of Non-Preemptive SJF P1P1 P3P3 P2P2 73160 P4P4 812

13. Silberschatz, Galvin, and Gagne  1999 6.13 Applied Operating System Concepts Example of Preemptive SJF (Also called Shortest-Remaining-Time-First (SRTF) ) ProcessArrival TimeBurst Time P 1 0.07 P 2 2.04 P 3 4.01 P 4 5.04 SJF (preemptive) Average waiting time = (9 + 1 + 0 +2)/4 = 3 P1P1 P3P3 P2P2 42 11 0 P4P4 57 P2P2 P1P1 16 In order for a new arrival to preempt, its burst must be strictly less than current remaining time

14. Silberschatz, Galvin, and Gagne  1999 6.14 Applied Operating System Concepts Determining Length of Next CPU Burst Can only estimate the length. Can be done by using the length of previous CPU bursts, using exponential averaging.  n+1 =  t n + (1-  )  n  n stores past history t n is “recent history  is a weighting factor … recursive in  n

15. Silberschatz, Galvin, and Gagne  1999 6.15 Applied Operating System Concepts Examples of Exponential Averaging  =0 –  n+1 =  n –Recent history does not count.  =1 –  n+1 = t n –Only the actual last CPU burst counts. If we expand the formula, we get:  n+1 =  t n +(1 -  )  t n -1 + … +(1 -  ) j  t n -1 + … +(1 -  ) n=1 t n  0 Since both  and (1 -  ) are less than or equal to 1, each successive term has less weight than its predecessor.

16. Silberschatz, Galvin, and Gagne  1999 6.16 Applied Operating System Concepts Priority Scheduling A priority number (integer) is associated with each process The CPU is allocated to the process with the highest priority (smallest integer  highest priority). –Preemptive –nonpreemptive SJF is a priority scheduling where priority is the predicted next CPU burst time. Problem: Starvation – low priority processes may never execute. Solution: Aging – as time progresses increase the priority of the process - a “reward” for waiting in line a long time - let the old geezers move ahead!

17. Silberschatz, Galvin, and Gagne  1999 6.17 Applied Operating System Concepts Round Robin (RR) Each process gets a small unit of CPU time (time quantum), usually 10-100 milliseconds. After this time has elapsed, the process is preempted and added to the end of the ready queue RR is a preemptive algorithm. If there are n processes in the ready queue and the time quantum is q, then each process gets 1/n of the CPU time in chunks of at most q time units at once. No process waits more than (n-1)q time units. Performance –q large  FIFO –q small  q must be large with respect to context switch, otherwise overhead is too high. ==> “processor sharing” - user thinks it has its own processor running at 1/n speed (n processors)

18. Silberschatz, Galvin, and Gagne  1999 6.18 Applied Operating System Concepts Example: RR with Time Quantum = 20 ProcessBurst Time P 1 53 P 2 17 P 3 68 P 4 24 The Gantt chart is: Typically, higher average turnaround than SJF, but better response. P1P1 P2P2 P3P3 P4P4 P1P1 P3P3 P4P4 P1P1 P3P3 P3P3 02037577797117121134154162 FCFS is tie breaker Assume all arrive at 0 time in the order given.

19. Silberschatz, Galvin, and Gagne  1999 6.19 Applied Operating System Concepts How a Smaller Time Quantum Increases Context Switches Context switch overhead very critical for 3rd case - since overhead is independent of quanta time

20. Silberschatz, Galvin, and Gagne  1999 6.20 Applied Operating System Concepts Turnaround Time Varies With The Time Quantum TAT can be improved if most processes finish each burst in one quanta EX: if 3 processes each have burst of 10, then for q = 1, avg_TAT = 29, but for q = burst = 10, avg_TAT = 20. ==> design tip: tune quanta to average burst. No strict correlation of TAT and time quanta size - except for below

21. Silberschatz, Galvin, and Gagne  1999 6.21 Applied Operating System Concepts Multilevel Queue (no Feedback - see later) Ready queue is partitioned into separate queues: foreground queue (interactive) background queue (batch) Each queue has its own scheduling algorithm, foreground queue – RR background queue – FCFS Scheduling must be done between the queues. –Fixed priority scheduling; i.e., serve all from foreground then from background. All higher priority queues must be empty before given queue is processed. Possibility of starvation. Assigned queue is for life of the process. –Time slice between queues – each queue gets a certain amount of CPU time which it can schedule amongst its processes; i.e., 80% to foreground in RR –20% to background in FCFS

22. Silberschatz, Galvin, and Gagne  1999 6.22 Applied Operating System Concepts Multilevel Queue Scheduling

23. Silberschatz, Galvin, and Gagne  1999 6.23 Applied Operating System Concepts Multilevel Feedback Queue A process can move between the various queues; aging can be implemented this way. Multilevel-feedback-queue scheduler defined by the following parameters: –number of queues –scheduling algorithms for each queue –method used to determine when to upgrade a process –method used to determine when to demote a process –method used to determine which queue a process will enter when that process needs service

24. Silberschatz, Galvin, and Gagne  1999 6.24 Applied Operating System Concepts Multilevel Feedback Queues Fig 6.7 Higher priority queues pre-empt lower priority queues on new arrivals Queue 0 - High priority Queue 1 Queue 2 Low priority

25. Silberschatz, Galvin, and Gagne  1999 6.25 Applied Operating System Concepts Example of Multilevel Feedback Queue Three queues: –Q 0 – time quantum 8 milliseconds –Q 1 – time quantum 16 milliseconds –Q 2 – FCFS –Q 2 longest jobs, with lowest priority, Q 1 shortest jobs with highest priority. Scheduling –A new job enters queue Q 0 which is served FCFS. When it gains CPU, job receives 8 milliseconds. If it does not finish in 8 milliseconds, job is moved to queue Q 1 (demoted). –At Q 1 job is again served FCFS and receives 16 additional milliseconds. If it still does not complete, it is preempted and moved to queue Q 2. –Again, Q n is not served until Q n-1 empty

26. Silberschatz, Galvin, and Gagne  1999 6.26 Applied Operating System Concepts Multiple-Processor Scheduling CPU scheduling more complex when multiple CPUs are available. Homogeneous processors within a multiprocessor. Load sharing Symmetric Multiprocessing (SMP) – each processor makes its own scheduling decisions. Asymmetric multiprocessing – only one processor accesses the system data structures, alleviating the need for data sharing.

27. Silberschatz, Galvin, and Gagne  1999 6.27 Applied Operating System Concepts Real-Time Scheduling Hard real-time systems – required to complete a critical task within a guaranteed amount of time. Soft real-time computing – requires that critical processes receive priority over less fortunate ones. Deadline scheduling used.

28. Silberschatz, Galvin, and Gagne  1999 6.28 Applied Operating System Concepts Dispatch Latency

29. Silberschatz, Galvin, and Gagne  1999 6.29 Applied Operating System Concepts Solaris 2 Thread Scheduling 3 scheduling priority classes –Timesharing/interactive – lowest – for users  Within this class: Multilevel feedback queues longer time slices in lower priority queues –System – for kernel processes –Real time – Highest Local Scheduling – How the threads library decides which thread to put onto an available LWP. Remember threads are time multiplexed on LWP’s Global Scheduling – How the kernel decides which kernel thread to run next. –The local schedules for each class are “globalized” from the scheduler's point of view – all classes included. –LPW’s scheduled by kernel

30. Silberschatz, Galvin, and Gagne  1999 6.30 Applied Operating System Concepts Solaris 2 Scheduling Fig. 6.10 Reserved for kernel use. Ex: Scheduler & paging daemon Only a few in this class Real time User processes Go here

31. Silberschatz, Galvin, and Gagne  1999 6.31 Applied Operating System Concepts Java Thread Scheduling JVM Uses a Preemptive, Priority-Based Scheduling Algorithm. FIFO Queue is Used if There Are Multiple Threads With the Same Priority.

32. Silberschatz, Galvin, and Gagne  1999 6.32 Applied Operating System Concepts Java Thread Scheduling (cont)-omit JVM Schedules a Thread to Run When: 1. The Currently Running Thread Exits the Runnable State. 2. A Higher Priority Thread Enters the Runnable State * Note – the JVM Does Not Specify Whether Threads are Time- Sliced or Not.

33. Silberschatz, Galvin, and Gagne  1999 6.33 Applied Operating System Concepts Time-Slicing (Java) - omit Since the JVM Doesn’t Ensure Time-Slicing, the yield() Method May Be Used: while (true) { // perform CPU-intensive task... Thread.yield(); } This Yields Control to Another Thread of Equal Priority. Cooperative multi-tasking possible using yield

34. Silberschatz, Galvin, and Gagne  1999 6.34 Applied Operating System Concepts Java Thread Priorities - omit Thread Priorities: PriorityComment Thread.MIN_PRIORITYMinimum Thread Priority Thread.MAX_PRIORITYMaximum Thread Priority Thread.NORM_PRIORITYDefault Thread Priority Priorities May Be Set Using setPriority() method: setPriority(Thread.NORM_PRIORITY + 2);

35. Silberschatz, Galvin, and Gagne  1999 6.35 Applied Operating System Concepts Algorithm Evaluation Deterministic modeling – takes a particular predetermined workload and defines the performance of each algorithm for that workload - what we’ve been doing in the examples - optimize various criteria. –Easy, but success depends on accuracy of input Queuing models - statistical - need field measurements of statistics in various compouting environments Implementation - Costly - OK is a lot of pre-implementation done first.

36. Silberschatz, Galvin, and Gagne  1999 6.36 Applied Operating System Concepts Evaluation of CPU Schedulers by Simulation -need good models - drive it with field data and/or statistical data - could be slow.