V0.1 CSS430 Scheduling Textbook Chapter 5 Instructor: Stephen G. Dame These slides were adapted from the OSC textbook slides (Silberschatz, Galvin, and Gagne), Professor Munehiro Fukuda and the instructor’s class materials. 1 CSS430 Operating Systems : Scheduling

V0.1 CPU Scheduling Learning Objectives 2 CSS430 Operating Systems : Scheduling  Basic Concepts  Scheduling Criteria  Scheduling Algorithms (FCFS, SJF, etc.)  Thread Scheduling  Multi-Processor Scheduling  OS Examples  Algorithm Evaluation

V0.1 TermDefinition PCBProcess Control Blocks FCFSFirst-Come First-Served scheduling algorithm SJFSmallest-Job-First scheduling algorithm RRRound Robin FIFOFirst In First Out (Queue) quantumsmallest schedulable time interval (“time slice”) Foregroundinteractive processes Backgroundbatch (or system) processes PCSprocess-contention scope SCSsystem-contention scope TATTurnaround TimeGlossary 3 CSS430 Operating Systems : Scheduling

V0.1 Processes vs. Threads 4 CSS430 Operating Systems : Scheduling  Based on last class... A fine distinction between processes and threads in Linux (e.g. tasks) A fine distinction between processes and threads in Linux (e.g. tasks)  Terms (in chapter 5) used interchangeably Process scheduling Process scheduling Thread scheduling Thread scheduling  Kernel level threads being scheduled – not processes.

V0.1 CPU Scheduling Goal 5 CSS430 Operating Systems : Scheduling The main objective of multiprogramming is to maximize CPU resource utilization for a given computer environment though optimal CPU scheduling. Example Computing Environments:   System Processes   Interactive processes   Batch oriented processes

V0.1 Histogram of CPU-burst Times 6 CSS430 Operating Systems : Scheduling CPU Burst DurationTYPE OF PROGRAM SHORTI/O Bound LONGCPU Bound

V0.1 CPU-I/O Burst Cycles 7 CSS430 Operating Systems : Scheduling CPU burst wait for I/O load store add store read from file I/O burst CPU burst wait for I/O store increment index write to file I/O burst CPU burst wait for I/O load store add store read from file I/O burst CPU burst wait for I/O load store add store read from file I/O burst CPU burst wait for I/O store increment index write to file I/O burst CPU burst wait for I/O load store add store read from file I/O burst Process A Process B Cycle n Cycle n+1 Cycle n+2

V0.1 CPU Scheduler 8 CSS430 Operating Systems : Scheduling Running Waiting Ready Exit UML Process State Diagram Wait I/O Complete Interrupt Ready Queue Short-term Scheduler If only and then scheduling = non-preemptive else scheduling = preemptive 14 Process Termination PCBs

V0.1 TermDefinition CPU Utilization Typically ranges from 40% to 90% - keep the CPU as busy as possible. ThroughputNumber of processes completed per unit of time. Turnaround Time (TAT) Time of process submission to time of completion. Waiting Time Total amount of time a process spends waiting in the ready queue. Response Time Time a process takes from submission to start of response. Scheduling Criteria 9 CSS430 Operating Systems : Scheduling

V0.1 First-Come First-Served (FCFS) 10 CSS430 Operating Systems : Scheduling  Example:ProcessBurst Time P 1 24 P 2 3 P 2 3 P 3 3 P 3 3  Suppose that the processes arrive in the order: P 1, P 2, P 3  Suppose that the processes arrive in the order: P 2, P 3, P 1.  Non-preemptive scheduling  Troublesome for time-sharing systems Convoy effect short process behind long process Convoy effect short process behind long process P1P1 P3P3 P2P P1P1 P2P2 P3P Waiting time for P 1 = 0; P 2 = 24; P 3 = 27 Average waiting time: ( )/3 = 17 Waiting time for P 1 = 6; P 2 = 0 ; P 3 = 3 Average waiting time: ( )/3 = 3 The process that requests the CPU first is allocated the CPU first.

V CSS430 Operating Systems : Scheduling 2 P 4 (2) (SJF) Average waiting time = ( )/4 = 3.75 P 2 (5) P 3 (0) P 1 (9) P 3 ‘s waiting time = 0 P 4 ‘s waiting time = 2 P 2 ‘s waiting time = 5 P 1 ‘s waiting time = 9 Proces s CPU Burst Time P17 P24 P32 P43 P1P1 P2P2 P3P3 P4P4 5 (FCFS) Average waiting time = ( )/4 = Shortest-Job-First (SJF) Non-preemptive

V0.1 Shortest-Job-First (SJF) Preemptive 12 CSS430 Operating Systems : Scheduling 2 P2(2)P2(2) (SJF) Average waiting time = ( )/4 = 3 P3(4)P3(4) P 1 (0) P4(6)P4(6) P 3 ‘s waiting time = 0 P 2 ‘s waiting time = 2 P 3 ‘s waiting time = 4 P 4 ‘s waiting time = 6 P1P1 P2P2 P3P3 P4P4 6 (FCFS) Average waiting time = ( )/4 = Proces s Arrival Time CPU Burst Time P107 P226 P342 P453 4 ~P 2 ~P 1 18

V Compute average wait time (FCFS) for all unique permutations of P1, P2, P3 where: 2. What is the difference between preemptive and non-preemptive schedulers? 3.Can an OS satisfy all the scheduling criteria: CPU utilization, throughput, TAT, waiting time, and response time? Discussion 1 13 CSS430 Operating Systems : Scheduling ProcessCPU Burst Time (msec) P134 P216 P337

V0.1 Shortest-Job-First (SJF) 14 CSS430 Operating Systems : Scheduling  Optimal scheduling  Preemptive SJF is superior to non-preemptive SJF.  However, there are no accurate estimations to know the length of the next CPU burst  Estimation in Figure 5.3 (recursive smoothing filter): Exponential Averaging Filter (recursive smoothing filter)

V0.1 SJF Predicted CPU Burst Example (Random Data) 15 CSS430 Operating Systems : Scheduling

V0.1 Priority Scheduling 16 CSS430 Operating Systems : 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 in Unix but lowest in Java). Preemptive Preemptive Non-preemptive Non-preemptive  SJF is a special case of priority scheduling priority is the predicted next CPU burst time. 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.

V CSS430 Operating Systems : Scheduling 1 P 5 (1) (Priority) Average waiting time = ( )/5 = 5.6 P 1 (6) P 3 (0) P 3 (9) Process CPU Burst Time Priority P173 P211 P334 P415 P552 P3P3 P1P1 P2P2 P5P5 6 (FCFS) Average waiting time = ( )/5 = Priority Scheduling Non-preemptive P4P P 4 (12)

V0.1 Round Robin Scheduling 18 CSS430 Operating Systems : Scheduling  Each process is given CPU time in turn, (i.e. time quantum: usually ms) ...thus waits no longer than ( n – 1 ) * time quantum  time quantum = 20ms P1P1 P2P2 P3P3 P4P4 P1P1 P3P3 P4P4 P1P1 P3P3 P3P P 1 (53) P 2 (17) P 3 (68) P 4 (24) P 1 (33) P 1 (13) P 3 (48) P 3 (28) P 3 (8) P 4 (4) 57 Process CPU Burst Time (ms) Wait Time (ms) TAT (ms) P = P P = P = AVERAGE41ms73ms114ms

V0.1 Round Robin Scheduling 19 CSS430 Operating Systems : Scheduling  Typically, higher average turnaround than SJF, but better response.  Performance (q = quantum) q large  FCFS q large  FCFS q small  q must be large with respect to context switch, otherwise overhead is too high. q small  q must be large with respect to context switch, otherwise overhead is too high.

V0.1 Turnaround Time (TAT) Varies With The Time Quantum 20 CSS430 Operating Systems : Scheduling TAT can be improved if most processes finish their next CPU burst in a single time quantum.

V0.1 Multilevel Queue Scheduling fixed distribution based on priority 21 CSS430 Operating Systems : Scheduling Each queue has its own scheduling algorithm, foreground (interactive) – RR, 80% CPU time background (batch) – FCFS, 20% CPU time

V0.1 Multilevel Feedback-Queue Scheduling (MFQS) – Variable based on consumption 22 CSS430 Operating Systems : Scheduling A new job enters queue Q 0 which is scheduled via FCFS. When it gains CPU, the current job receives 8 milliseconds. If it does not finish in 8 milliseconds, the job is moved to queue Q 1. At Q 1 job is again scheduled via FCFS and receives 16 additional milliseconds. If it still does not complete, it is preempted and moved to queue Q 2.

V0.1 Contention Scope 23 CSS430 Operating Systems : Scheduling TermDefinition PCSMany-to-many and many-to-one thread models schedule user-level threads to run on an available LWP SCSMapping kernel threads onto a CPU NOTE: Linux, Windows only use SCS because they are one-to-one thread models.

V0.1 Discussions 2 24 CSS430 Operating Systems : Scheduling ① What is the main problem in MFQS? ② How can you address this problem? Briefly think about your own personal scheduling algorithm?

V0.1 Cooperative multithreading in Java 25 CSS430 Operating Systems : Scheduling  A thread voluntary yielding control of the CPU is called cooperative multitasking.  Thread.yield( ) is OS-dependent. Linux does not care about it. Linux does not care about it. Solaris allows a thread to yield its execution when the corresponding LWP exhausts a given time quantum. Solaris allows a thread to yield its execution when the corresponding LWP exhausts a given time quantum.  Conclusion: Don’t write programs based on yield( ).

V0.1 Java-Based Round-Robin Scheduler (Naïve Example) 26 CSS430 Operating Systems : Scheduling

V0.1 Why does this not work? 27 CSS430 Operating Systems : Scheduling  Java: runs threads with a higher priority until they are blocked. (Threads with a lower priority cannot run concurrently.) This may cause starvation or deadlock. This may cause starvation or deadlock.  Java’s strict priority scheduling was implemented in Java green threads.  In Java 2, thread’s priority is typically mapped to the underlying OS-native thread’s priority.  OS: gives more CPU time to threads with a higher priority. (Threads with a lower priority can still run concurrently.)  If we do not want to depend on OS scheduling, we have to use: Thread.suspend( ) Thread.suspend( ) Thread.resume( ) Thread.resume( )

V0.1 Multiprocessor Scheduling 28 CSS430 Operating Systems : Scheduling  OS code and user process mapping: Asymmetric multiprocessing Asymmetric multiprocessing Symmetric multiprocessing (SMP) Symmetric multiprocessing (SMP)  Scheduling criteria: Processor affinity Processor affinity Load balancing Load balancing  Thread context switch: Coarse-grained software thread Coarse-grained software thread Fine-grained hardware thread Fine-grained hardware thread

V0.1 Algorithm Evaluation 29 CSS430 Operating Systems : Scheduling  Define the selection criteria (i.e.): Optimize CPU Utilization OR Optimize CPU Utilization OR Maximize throughput Maximize throughput  Evaluation Methods Deterministic Modeling (analytic method) Deterministic Modeling (analytic method) Queueing Models (see Little’s formula next) Queueing Models (see Little’s formula next) Simulations Simulations Implementation Implementation

V0.1 Little’s Formula 30 CSS430 Operating Systems : Scheduling n Average queue length Average arrival rate for new processes in queue W Average waiting time in queue

V0.1 Little’s Formula - Examples 31 CSS430 Operating Systems : Scheduling

V0.1Exercises 32 CSS430 Operating Systems : Scheduling ① Program #2: Check the syllabus for its due date ①(turn-in) Exercise Build a table or spreadsheet....for next Monday midnight deadline Note: this is in addition to the Tuesday assignment.