1. Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition, Chapter 5: CPU Scheduling

2. 5.2 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Chapter 5: CPU Scheduling Basic Concepts Scheduling Criteria Scheduling Algorithms Thread Scheduling Multiple-Processor Scheduling Operating Systems Examples Algorithm Evaluation

3. 5.3 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Objectives To introduce CPU scheduling, which is the basis for multiprogrammed operating systems To describe various CPU-scheduling algorithms To discuss evaluation criteria for selecting a CPU-scheduling algorithm for a particular system

4. 5.4 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Basic Concepts Maximum CPU utilization obtained with multiprogramming CPU–I/O Burst Cycle – Process execution consists of a cycle of CPU execution and I/O wait While a process waits for I/O, CPU sits idle if no multiprogramming Instead the OS can give CPU to another process CPU burst distribution

5. 5.5 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Alternating Sequence of CPU And I/O Bursts

6. 5.6 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Histogram of CPU-burst Times

7. 5.7 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition CPU Scheduler Short-term 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.Switches from running to waiting state 2.Switches from running to ready state 3.Switches from waiting to ready 4.Terminates Scheduling under 1 and 4 is nonpreemptive/cooperative All other scheduling is preemptive

8. 5.8 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition CPU Scheduler Nonpreemptive: Once the process is allocated the CPU, it keeps it until termination/wait. eg. Windows 3.x/95 No special hardware (like timers) needed. Preemptive scheduling – running process can be removed for another Issues: Shared data consistency – Synchronization (Ch. 6) What happens when the kernel is in a system call and the process asking for that call is preempted? UNIX – context switches can only happen after system calls. Other solutions – Sec 5.5, 19.5 Typically we cannot disable interrupts

9. 5.9 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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

10. 5.10 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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 Waiting time – amount of time a process has been waiting in the ready queue 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)

11. 5.11 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Scheduling Algorithm Optimization Criteria Max CPU utilization Max throughput Min turnaround time Min waiting time Min response time

12. 5.12 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Scheduling Algorithms First-Come, First-Served Scheduling Shortest-Job-First Scheduling Priority Scheduling Round-Robin Scheduling Multilevel Queue Scheduling Multilevel Feedback Queue Scheduling

13. 5.13 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition First-Come, First-Served (FCFS) Scheduling 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

14. 5.14 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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 – open CPU bound process followed by multiple I/O processes. Nonpreemptive P1P1 P3P3 P2P2 63300

15. 5.15 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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 If burst times are the same – break ties using FCFS SJF is provably optimal – gives minimum average waiting time for a given set of processes Reasoning – move the short process before a long one. This decreases the waiting time of the short process more than it increases the waiting time of the long one. Hence the average decreases. The difficulty is knowing the length of the next CPU request

16. 5.16 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Example of SJF ProcessArrival TimeBurst Time P 1 0.06 P 2 2.08 P 3 4.07 P 4 5.03 SJF scheduling chart Average waiting time = (3 + 16 + 9 + 0) / 4 = 7 P4P4 P3P3 P1P1 3 16 0 9 P2P2 24

17. 5.17 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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

18. 5.18 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Prediction of the Length of the Next CPU Burst

19. 5.19 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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 -j + … +(1 -  ) n +1  0 Since both  and (1 -  ) are less than or equal to 1, each successive term has less weight than its predecessor Can be preemptive or nonpreemptive

20. 5.20 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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

21. 5.21 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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. 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. Also called processor sharing – appears like each process has a processor

22. 5.22 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Example of RR with Time Quantum = 4 ProcessBurst Time P 1 24 P 2 3 P 3 3 The Gantt chart is: Typically, higher average turnaround than SJF, but better response Avg Wait=17/3=5.66 P1P1 P2P2 P3P3 P1P1 P1P1 P1P1 P1P1 P1P1 0 47 101418222630

23. 5.23 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Time Quantum and Context Switch Time

24. 5.24 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Turnaround Time Varies With The Time Quantum

25. 5.25 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multilevel Queue Ready queue is partitioned into separate queues: foreground (interactive) background (batch) Each queue has its own scheduling algorithm foreground – RR background – FCFS Scheduling must be done between the queues Fixed priority scheduling; (i.e., serve all from foreground then from background). Possibility of starvation. Time slice – 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

26. 5.26 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multilevel Queue Scheduling

27. 5.27 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition 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

28. 5.28 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Example of Multilevel Feedback Queue Three queues: Q 0 – RR with time quantum 8 milliseconds Q 1 – RR time quantum 16 milliseconds Q 2 – FCFS 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. 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.

29. 5.29 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multilevel Feedback Queues

30. 5.30 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Thread Scheduling Distinction between user-level and kernel-level threads OS only schedules kernel-level threads. User-level threads are scheduled through a direct or indirect (LWP) mapping Many-to-one and many-to-many models, thread library schedules user-level threads to run on LWP Known as process-contention scope (PCS) since scheduling competition is within the process Kernel thread scheduled onto available CPU is system-contention scope (SCS) – competition among all threads in system Typically – PCS is priority based. Programmer can set user-level thread priorities

31. 5.31 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Pthread Scheduling API allows specifying either PCS or SCS during thread creation PTHREAD_SCOPE_PROCESS schedules threads using PCS scheduling PTHREAD_SCOPE_SYSTEM schedules threads using SCS scheduling. 2 methods to get and set the scope pthread_attr_setscope(pthread_attr_t *attr, int scope) pthread_attr_getscope(pthread_attr_t *attr, int *scope) attr – pointer to the attribute set scope – PSC/SCS

32. 5.32 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Pthread Scheduling API #include #define NUM THREADS 5 int main(int argc, char *argv[]) { int i; pthread_t tid[NUM THREADS]; pthread_attr_t attr; /* get the default attributes */ pthread_attr init(&attr); /* set the scheduling algorithm to PROCESS or SYSTEM */ pthread_attr setscope(&attr, PTHREAD_SCOPE_SYSTEM); /* set the scheduling policy - FIFO, RT, or OTHER */ pthread_attr setschedpolicy(&attr, SCHED_OTHER); /* create the threads */ for (i = 0; i < NUM THREADS; i++) pthread create(&tid[i],&attr,runner,NULL);

33. 5.33 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Pthread Scheduling API /* now join on each thread */ for (i = 0; i < NUM THREADS; i++) pthread join(tid[i], NULL); } /* Each thread will begin control in this function */ void *runner(void *param) { printf("I am a thread\n"); pthread exit(0); }

34. 5.34 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multiple-Processor Scheduling CPU scheduling more complex when multiple CPUs are available ASSUMPTION - Homogeneous processors within a multiprocessor Asymmetric multiprocessing – only one processor accesses the system data structures, alleviating the need for data sharing Problems – Bottleneck, Single point of failure Symmetric multiprocessing (SMP) – each processor is self-scheduling, all processes in common ready queue, or each has its own private queue of ready processes\ Most common – Windows XP, 2000, Linux, OS X Remainder of the discussion applies to SMP.

35. 5.35 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multiprocessor Scheduling Processor affinity – process has affinity for processor on which it is currently running Reason – caching. As a process runs, the cache gets populated and it is increasingly likely that the requests will be satisfied from the cache. soft affinity OS tries to keep the process running on the same processor, but this is not binding. Migration possible. hard affinity Supported in Linux. Allows a process to specify this. Solaris supports the creation of processor sets. Also soft but somewhat more restricted.

36. 5.36 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multiprocessor Scheduling Affinity may be decided by the architecture of the main-memory. NUMA – Non Uniform Memory Access CPU has faster access to some memory. Multiprocessors systems where each CPU has a memory board. It can also access memory on other CPU’s but there is a delay OS design influenced by the architecture and optimized for performance

37. 5.37 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition NUMA and CPU Scheduling

38. 5.38 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Load Balancing Goal: Keep workload evenly distributed between processors Usually only necessary if each processor has its own individual queue If there is a common queue, a processor can just pick a job from here when free Push migration – specific task to check load on each processor and redistribute if needed. Pull migration – idle processor pulls task form a busy one. Usually both are implemented in parallel. Eg. Linux scheduler Note that this conteracts with affinity.

39. 5.39 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multicore Processors Recent trend to place multiple processor cores on same physical chip Each core has its own register set Faster and consume less power Multiple threads per core also growing Takes advantage of memory stall to make progress on another thread while memory retrieve happens Memory stalls – can be upto 50% of the time Solution – Multithreaded Multicores

40. 5.40 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multithreaded Multicore System

41. 5.41 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Multithreaded Multicore Coarse-grained More cost with switching between threads Fine-grained Much finer level of granularity in switching between threads – logic for thread switching included in architecture 2 levels of scheduling are happening here.

42. 5.42 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Operating System Examples Solaris scheduling Windows XP scheduling Linux scheduling

43. 5.43 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Solaris Scheduling Priority based thread scheduling with 6 classes: Time sharing Interactive Real Time System Fair Share Fixed Priority Default class for a process is Time Sharing TS – dynamic priorities and slice lengths using a multilevel queue Eg. Shown for different priorities

44. 5.44 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Solaris Dispatch Table

45. 5.45 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Solaris Dispatch Table Priority – Higher number, Higher Priority Quantum length inversely proportional to Priority Time Quantum expired – new priority for thread that has used its entire quantum without blocking (CPU intensive threads) Return from sleep – Priority of a thread returning from a sleep state eg. Waiting for I/O. When I/O is available its priority is boosted

46. 5.46 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Solaris Scheduling

47. 5.47 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Windows XP Scheduling 32-level priority scheme to determine order of execution Split into two classes Variable class – 1-15 Real-time class – 16-31 Several Priority classes in the API, followed by relative priority within a class Variable priorities Typically priority of the foreground process is increased – usually by 3

48. 5.48 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Windows XP Priorities

49. 5.49 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Linux Scheduling Constant order O(1) scheduling time Regardless of number of tasks Two priority ranges: time-sharing(nice) and real-time Real-time range from 0 to 99 and nice value from 100 to 140 Lower values -> Higher priorities Unlike Solaris, Higher priority is given Larger Time slice

50. 5.50 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Priorities and Time-slice length

51. 5.51 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition List of Tasks Indexed According to Priorities

52. 5.52 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Algorithm Evaluation Deterministic modeling – takes a particular predetermined workload and defines the performance of each algorithm for that workload eg. Calculating the average wait time for each model Queueing models Implementation

53. 5.53 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Analytical Evaluation Processes Burst Time P110 P229 P33 P47 P512

54. 5.54 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition FCFS Wait time=(0+10+39+42+49)/5=28ms SJF Wait time=(10+32+0+3+20)/5=13 ms RR Wait time=(0+32+20+23+40)/5=23 ms

55. 5.55 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Queuing Network Analysis Processes are dynamic and cannot be estimated But CPU and I/O burst distributions can be Formula estimating the probability of a particular burst Similarly arrival times can be shown by a distribution Given these two distributions, possible to compute avg throughput, utilization, waiting time etc. Let n : avg. queue length W : avg wait time in the queue : Avg Arrival Rate n = x W Little’s Formula If the system is steady, number entering must be equal to number leaving

56. 5.56 Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition Evaluation of CPU schedulers by Simulation

57. Silberschatz, Galvin and Gagne ©2009 Operating System Concepts – 8 th Edition, End of Chapter 5