Chap. 6: CPU Scheduling

Transcription

1 Chap. 6: CPU Scheduling FBasic Concepts FScheduling Criteria FScheduling Algorithms FMultiple-Processor Scheduling FReal-Time Scheduling FThread Scheduling FJava Thread Scheduling FAlgorithm Evaluation 1 CPU and I/O Burst load store add store read from file wait for I/O store increment index write to file wait for I/O load store add store read from file wait for I/O.. CPU burst I/O burst CPU burst I/O burst CPU burst I/O burst Frequency Burst duration (milliseconds) 2 Page 1

2 CPU Scheduling FThe task of selecting a waiting process from the ready queue and allocating the CPU to it. FCPU scheduling decision may take place under the following four conditions: (1) When a process switches from the running state to the waiting state. (2) When a process switches from the running state to the ready state. (3) When a process switches from the waiting state to the ready state. (4) When a process terminates. 3 Preemptive Scheduling FNonpreemptive scheduling CPU scheduling takes place only under conditions (1) and (4). That is» When a process switches from the running state to the waiting state.» When a process terminates FPreemptive scheduling: Otherwise processes may be preempted while updating some shared data 4 Page 2

3 Process State Diagram New Admitted Interrupt (2) (4) Terminated Exit Preemptive Ready Running (3) I/O or event completion Waiting Non-Preemptive (1) I/O or event wait 5 Dispatcher FThe module that gives control of CPU to the process selected by the short-term scheduler. FThis function involves: Switching context Switching to user mode Jumping to the proper location in the user program to restart that program FDispatch latency: the time it takes to stop one process and start another running process. 6 Page 3

4 Scheduling Criteria FFor comparing CPU scheduling algorithms CPU Utilization Throughput (number of completed processes per time unit) Turnaround Time (time from submission to complete) Waiting Time (total waiting time in the ready queue) Response Time (submission of a request until the first response) 7 Scheduling Algorithms FFirst-Come, First-Served Scheduling FShortest-Job-First Scheduling FPriority Scheduling FRound-Robin Scheduling FMultilevel Queue Scheduling FMultilevel Feedback Queue Scheduling 8 Page 4

5 First-Come, First-Served Scheduling FNon-preemptive. Has a long average waiting time. FFor example: Process Burst Time P1 24 P2 3 P3 3 P1 P2 P P2 P AWT = ( )/3 = 17 FConvoy effect: all the other processes wait for the one big process to get off the CPU. P1 AWT = (6+0+3)/3 = 3 9 The Convoy Effect The ready queue IO-bound IO-bound IO-bound CPUbound CPU The ready queue IO-bound IO-bound CPUbound CPU IO-bound IO-bound IO-bound CPU IO-bound I/O I/O queue CPUbound I/O 10 Page 5

6 Shortest-Job First Scheduling (SJF) FThe smallest next CPU burst is selected. FProvides the minimum average waiting time (optimal). FFrequently used in long-term scheduling (Not short-term ). FThe next CPU burst can be predicted as an exponential average of the measured length of previous CPU bursts. P4 P1 Process P3 Burst Time P1 6 P2 8 P3 P P2 AWT = ( )/4 = 7 11 Predicting CPU Burst Burst length CPU burst Guess t n+1: predicted value t n : length of the nth CPU burst t n+1 = a t n +(1-a) t n a =1/2 Time 12 Page 6

7 Preemptive SJF Scheduling FWhen a new process arrives the ready queue while a previous process is executing. The new process may have a shorter next CPU burst A preemptive SJF algorithm will preempts the current executing one» sometimes called shortest-remaining-time-first scheduling. 13 SJF Scheduling Examples Process Arrival Time Burst Time P1 0 8 P2 1 4 P3 P P1 P2 P4 P1 P3 Preemptive AWT = ((10-1)+(1-1)+(17-2)+(5-3))/4 = 6.5 P1 P AWT = ((0-0)+(8-1)+(17-2)+(12-3))/4 = 7.75 P4 P3 FThe SJF may be either preemptive or nonpreemptive. Non- Preemptive 14 Page 7

8 Priority Scheduling FThe highest priority process is selected first. FInternally or externally defined FCan be either preemptive or non-preemptive. FMay have indefinite blocking or starvation (aging can be used). Process Burst Time Priority P P2 1 1 P3 2 4 P4 1 5 P5 5 2 Low numbers represent high priority P2 P5 P1 P3 P Round-Robin (RR) Scheduling FDesigned specially for time-sharing systems. FSimilar to FCFS, but preemption is added. FA time quantum (time slice) is defined ( ms). FOften provides a long average waiting time. FPerformance depends heavily on the size of the time slice. Process Burst Time P1 24 P2 3 P3 3 AWT = 17/3 = 5.67 P1 P2 P3 P1 P1 P1 P1 P Page 8

9 RR Scheduling Properties Avg. turnaround time P time P1 P2 6 3 P3 1 P4 7 Process time = q context switches Smaller quantum increases context switches Time quantum 17 Multilevel Queue Scheduling FProcesses are classified into different groups different groups have different response-time requirements Fthe ready queue is partitioned into several queues each queue has its own scheduling algorithm processes are permanently assigned into one queue scheduling between queues» fixed-priority preemptive scheduling» time slice between the queues 18 Page 9

10 Multilevel Queue Scheduling Example Highest priority System processes Interactive processes Interactive editing processes Batch processes Student processes In case of fixed-priority preemptive scheduling, no process in batch queue can be run unless all higher-priority queues are empty Lowest priority 19 Multilevel Feedback Queue Scheduling FProcesses are allowed to move between queues. FSeparate processes with different CPU-burst characteristics. If a process uses too much CPU time, it will be moved to a lower-priority queue.» leaves I/O-bound and interactive processes in the highpriority queue A process that waits too long in a lower-priority queue may be moved to a higher-priority queue» to prevent starvation 20 Page 1 0

11 Multilevel Feedback Queue Scheduling quantum = 8 quantum = 24 Gives the highest priority to any process with a CPU burst of 8 ms or less. FCFS 21 Multilevel Feedback Queue Scheduling Fdefined by the following parameters: The number of queues The scheduling algorithm for each queue The method used to determine when to upgrade a process to a higher-priority queue The method used to determine when to demote a process to a lower-priority queue The method used to determine which queue a process will enter when that process needs service FThe most general CPU scheduling algorithm. 22 Page 1 1

12 Multiple-Processor Scheduling FHomogeneous and Heterogeneous systems. FLoad sharing can be done for several identical processors. FSeparate queue scheme vs a common ready queue scheme. FAll processes go into one queue and are scheduled onto any available processor. Two scheduling approaches: Each processor is self-scheduling (carefully designed). Appoint one processor as scheduler for the other processes. The Master-slave structure. Fasymmetric multiprocessing - all system activities are handled by one specific processor. Others only execute user code. 23 Real-Time Computing FHard real-time systems required to complete a critical task within a guaranteed amount of time. resource reservation is needed for admission control. FSoft real-time computing requires that critical processes receive priority over less fortunate ones. may result in longer delay, or even starvation, for some processes could support multimedia tasks 24 Page 1 2

13 Implementing Soft Real- Time Functionality FImplementation issues of soft real-time functions The system must have priority scheduling» the real-time process must have the highest priority which must not degrade over time. (no aging) The dispatch latency must be small. To do this, the system calls should be pre-emptible.» Insert preemptive points in long-duration system calls» make the entire kernel preemptible (Solaris 2) 25 Preemptible Kernel FWhat happens if the higher-priority process needs to read or modify kernel data that are currently being accessed by a lower-priority process? The higher-priority process would be waiting for a lower-priority one to finish (priority inversion) priority inversion FPriority inheritance protocol: all processes accessing resources that the high-priority process needs inherit the high priority until they are done with the resource in question. 26 Page 1 3

14 Dispatch Latency Event Interrupt processing Process made available Response interval Dispatch latency Response to event Preempt processes & release resources Real-time process execution Conflicts Dispatch time 27 Conflict Phase FThe conflict phase of dispatch latency has two components: Preemption of any process running in the kernel Release by low-priority processes of resources needed by the higher-priority process FIn Solaris 2 the dispatch latency with preemption disabled is over 100 ms. When enabled, it takes only 2 ms!! 28 Page 1 4

15 Thread Scheduling FLocal Scheduling How the threads library decides which thread to put onto an available LWP FGlobal Scheduling How the kernel decides which kernel thread to run next 29 Four classes of scheduling real time system time sharing interactive There is a set of priorities within each scheduling class Solaris 2 Scheduling 30 Page 1 5

16 Java Thread Scheduling FJVM uses a preemptive, priority-based scheduling algorithm. FIFO queue is used if there are multiple threads with the same priority. FJVM schedules a thread to run when: the currently running thread exits the runnable state. a higher priority thread enters the runnable state * Note the JVM does not specify whether threads are time-sliced or not. 31 Time-Slicing FSince 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. 32 Page 1 6

17 Thread Priorities FThread Priorities: Priority Comment Value Thread.MIN_PRIORITY Minimum Thread Priority 1 Thread.MAX_PRIORITY Maximum Thread Priority 10 Thread.NORM_PRIORITY Default Thread Priority 5 Priorities may be set using setpriority() method: setpriority(thread.norm_priority + 2); 33 Algorithm Evaluation FCriteria to select a CPU scheduling algorithm may include several measures, such as : Maximize CPU utilization under the constraint that the maximum response time is 1 second Maximize throughput such that turnaround time is (on average) linearly proportional to total execution time FHow to evaluate a selected algorithm? Analytic evaluation» Deterministic modeling» Queuing models Simulations Implementation 34 Page 1 7

18 Deterministic Modeling Ftakes a particular predetermined workload and defines the performance of each algorithm for that workload. simple and fast gives the exact numbers, allows the algorithms to be compared. requires exact numbers of input, and its answers apply to only those cases. FIn general, deterministic modeling is too specific, and requires too much exact knowledge, to be useful. 35 Deterministic Modeling Example Process Burst Time P1 10 P2 29 P3 3 P4 7 P5 12 FCFS P1 P2 P3 P4 P AWT = 28 ms SJF P3 P4 P1 P5 P AWT = 13 ms RR (q=10) P1 P2 P3 P4 P5 P2 P5 P AWT = 23 ms 36 Page 1 8

19 Queuing Models FProcesses that are run on many systems vary from day to day there is no static set of processes (and times) to use for deterministic modeling. FWhat can be determine the distribution of CPU and I/O bursts» a mathematical formula describing the probability of a particular CPU burst. the distribution of processes arrival time FThese distribution may be measured and then approximated or simply estimated. 37 Modeling Computer System FThe computer system is described as a network of servers. Each server has a queue of waiting processes. The CPU is a server with its ready queue. FQueuing-network analysis known» arrival rates and service rates to compute» utilization, average queue length, average waiting time, and so on. 38 Page 1 9

20 Queuing Analysis FLittle's formula: n : average queue length n = l x W when process arrival rate = process departure rate W: average waiting time in the queue l : average arrival rate for new processes We can use this formula to compute one of the three variables, if we know the other two. Fuseful in comparing scheduling algorithms, but has limitations. The arrival and service distributions are often defined in unrealistic, but mathematically tractable, ways. 39 Simulations FSimulations get a more accurate evaluation of scheduling algorithms. Fdistribution-driven simulation randomly generates data the distribution may be defined mathematically or empirically F simulation with trace tapes A trace tape can be created by monitoring the real system, recording the sequence of actual events. 40 Page 2 0

21 Simulations actual process execution record CPU 10 I/O 320 CPU 20 I/O 123 CPU 135 I/O 100 FCFS SJF RR (Q=14) Performance Statistics 41 Page 2 1