Transcript Module 6: CPU Scheduling - Simon Fraser University
Chapter 5: CPU Scheduling
Chapter 5: Objectives
Understand Scheduling Criteria Scheduling Algorithms Multiple-Processor Scheduling
Operating System Concepts 5.2
Silberschatz, Galvin and Gagne ©2005
Basic Concepts
Maximum CPU utilization obtained with multiprogramming CPU –I/O Burst Cycle Process execution consists of a of CPU execution and I/O wait
cycle
How long is the CPU burst?
Operating System Concepts 5.3
Silberschatz, Galvin and Gagne ©2005
CPU Burst Distribution
CPU bursts vary greatly from process to process and from computer to computer But, in general, they tend to have the following distribution (expo) Many short bursts Few long bursts
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.4
CPU Scheduler
Selects
one
process from
ready queue
to run on CPU Scheduling can be
Nonpreemptive
Once a process is allocated the CPU, it does
not
leave unless: 1. it has to wait, e.g., for I/O request or for a child to terminate 2. it terminates
Preemptive
OS can force (preempt) a process from CPU at
anytime
– Say, to allocate CPU to another higher-priority process Which is harder to implement? and why? Preemptive is harder: Need to maintain
consistency
of data shared between processes, and more importantly, kernel data structures (e.g., I/O queues) – Think of a preemption while kernel is executing a sys call on behalf of a process (many OSs, wait for sys call to finish)
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.5
Dispatcher
Scheduler
: selects one process to run on CPU
Dispatcher
: allocates CPU to the selected process, which involves: switching context switching to user mode jumping to the proper location (in the selected process) and restarting it
Dispatch latency
– time it takes for the dispatcher to stop one process and start another running
How does scheduler select a process to run?
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.6
Scheduling Criteria
CPU utilization
– keep the CPU as busy as possible Maximize
Throughput
– # of processes that complete their execution per time unit Maximize
Turnaround time
– amount of time to execute a particular process (time from submission to termination) Minimize
Waiting time
– amount of time a process has been waiting in the
ready queue
Minimize
Response time
– amount of time it takes from when a request was submitted until the first response is produced, sharing environment)
not
output (for time Minimize
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.7
Scheduling Algorithms
First Come, First Served Shortest Job First Priority Round Robin Multilevel queues Note: A process may have many CPU bursts, but in the examples we show only one for simplicity
5.8
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts
First-Come, First-Served (FCFS) Scheduling
Process Burst Time
P 1 P 2
24 3
P 3
3 Suppose that the processes arrive in the order:
P 1
The Gantt Chart for the schedule is: ,
P 2
,
P 3
P 1 P 2 P 3 27 30 0 24 Waiting time for
P 1
= 0;
P 2
= 24;
P 3
= 27 Average waiting time: (0 + 24 + 27)/3 = 17
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.9
FCFS Scheduling (Cont.)
Suppose that the processes arrive in the order
P 2
,
P 3
,
P 1
3, 3, 24 The Gantt chart for the schedule is: P 2 P 3 P 1 0 3 6 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 30
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.10
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
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.11
Example of Non-Preemptive SJF
Process
P 1 P 2 P 3 P 4
SJF (non-preemptive) Arrival Time 0.0
2.0
4.0
5.0
P 1 P 3 Burst Time 7 4 1 4 P 2 0 3 7 8 Average waiting time = (0 + 6 + 3 + 7)/4 = 4 12 P 4 16
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.12
Example of Preemptive SJF
Process Arrival Time
P 1 P 2 P 3 P 4
SJF (preemptive, SRJF) 0.0
2.0
4.0
5.0
P 1 P 2 P 3 P 2 P 4 0 2 4 5 7 11 Average waiting time = (9 + 1 + 0 +2)/4 = 3 Burst Time 7 4 1 4 P 1 16
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.13
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
1.
t n
2.
n
1
3.
, 0 actual lenght of
n th
CPU burst
predicted value for the next CPU burst
1 4.
Define :
n
1
t n
1
n
Operating System Concepts 5.14
Silberschatz, Galvin and Gagne ©2005
Exponential Averaging
If we expand the formula, we get:
n
+1 =
t
n
+
(1
-
) t
n
-1 +
(
1 -
)
2
t
n
-
2
+ …
+ (
1 -
)
n
+1
0
Since both and (1 ) are less than or equal to 1, each successive term has less weight than its predecessor Examples: = 0 ==> n+1 value) = n ==> Last CPU burst does not count (transient =1 ==> n+1 =
t
n ==> Only last CPU burst counts (history is stale)
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.15
Prediction CPU Burst Lengths: Expo Average
Assume = 0.5, 0 = 10
Operating System Concepts 5.16
Silberschatz, Galvin and Gagne ©2005
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 Aging: increase the priority of a process as it waits in the system
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.17
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 FCFS
q
small
q
must be large with respect to context switch, otherwise overhead is too high
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.18
Example of RR with Time Quantum = 20
Process
P 1 P 2 P 3 P 4
The Gantt chart is: Burst Time 53 17 68 24 0 P 1 20 P 2 37 P 3 57 P 4 77 P 1 P 3 97 117 P 4 P 1 P 3 P 3 121 134 154 162 Typically, higher average turnaround than SJF, but better response
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.19
Time Quantum and Context Switch Time
Smaller q more responsive but more context switches (overhead)
Operating System Concepts 5.20
Silberschatz, Galvin and Gagne ©2005
Turnaround Time Varies With The Time Quantum
Turnaround time varies with quantum, then stabilizes Rule of thumb for good performance: 80% of CPU bursts should be shorter than time quantum
Operating System Concepts 5.21
Silberschatz, Galvin and Gagne ©2005
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; e.g., 80% to foreground in RR 20% to background in FCFS
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.22
Multilevel Queue Scheduling
Operating System Concepts 5.23
Silberschatz, Galvin and Gagne ©2005
Multilevel Feedback Queue
A process
can move
implemented this way between various queues; aging can be 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
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.24
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
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.25
Multilevel Feedback Queues
Notes: Short processes get served faster (higher prio) more responsive Long processes (CPU bound) sink to bottom served FCFS more throughput
Operating System Concepts 5.26
Silberschatz, Galvin and Gagne ©2005
Multiple-Processor Scheduling
Multiple processors ==> divide load among them More complex than single CPU scheduling How to divide load?
Asymmetric multiprocessor One master processor does the scheduling for others Symmetric multiprocessor (SMP) Each processor runs its own scheduler One common ready queue for all processors, or one ready queue for each Win XP, Linux, Solaris, Mac OS X support SMP
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.27
SMP Issues
Processor affinity When a process runs on a processor, some data is cached in that processor’s cache A process migrates to another processor ==> Cache of new processor has to be
re-populated
Cache of old processor has to be
invalidated
==> performance penalty Load balancing One processor has too much load and another is idle Balance load using
Push migration
: A specific task periodically checks load on all processors and evenly distributes it by moving (pushing) tasks
Pull migration
: Idle processor pulls a waiting task from a busy processor Some systems (e.g., Linux) implement both Tradeoff between load balancing and processor affinity: what would you do?
May be, invoke load balancer when imbalance exceeds a threshold
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.28
Real-time Scheduling
Hard-real time systems A task must be finished within a deadline Ex: Control of spacecraft Soft-real time systems A task is given higher priority over others Ex: Multimedia systems
Operating System Concepts 5.29
Silberschatz, Galvin and Gagne ©2005
Operating System Examples
Windows XP scheduling Linux scheduling
Operating System Concepts 5.30
Silberschatz, Galvin and Gagne ©2005
Windows XP Scheduler
Priority-based, preemptive scheduler The highest-priority thread will always run 32 levels of priorities, each has a separate queue Scheduler traverses queues from highest to lowest till it finds a thread that is ready to run Priorities are divided into classes, each has several
relative
priorities
Operating System Concepts 5.31
Silberschatz, Galvin and Gagne ©2005
Windows XP Scheduler (cont’d)
Real-time class (fixed):
Levels 16 to 31
Other classes (variable):
Levels 1 to 15 Priority may change (decrease or increase) Priority decreases After thread’s quantum time runs out but never goes below the base (normal) value of its class Limit CPU consumption of CPU-bound threads Priority increases After a thread is released from a wait operation Bigger increase if thread was waiting for mouse or keyboard Moderate increase if it was waiting for disk Also, active window gets a priority boost Yield good response time
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.32
Linux Scheduler
Priority-based, preemptive scheduler with two separate ranges Real-time: 0 to 99 Nice: 100 to 140 (from -20 to 19) Higher priority tasks get larger quanta (unlike Win XP, Solaris)
Operating System Concepts 5.33
Silberschatz, Galvin and Gagne ©2005
Linux Scheduler (cont’d)
A task is initially assigned a time slice (quantum) Runqueue has two arrays:
active
and
expired
A runnable task is eligible for CPU if it still has time left in its time slice If the time slice runs out, the task is moved to the expired array When there are no tasks in the active array, the expired array becomes the active array and vice versa (change of pointers)
Operating System Concepts 5.34
Silberschatz, Galvin and Gagne ©2005
Algorithm Evaluation
Deterministic modeling Takes a particular predetermined workload and defines the performance of each algorithm for that workload Not general Queuing models Use queuing theory to analyze algorithms Many (unrealistic) assumptions to facilitate analysis Simulation Build a simulator and test with synthetic workload (e.g., generated randomly), or Traces collected from running systems Implementation Code it up and test!
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.35
Summary
Process execution: cycle of CPU bursts and I/O bursts CPU bursts lengths: many short bursts, and few long ones Scheduler selects one process from ready queue Dispatcher performs the switching Scheduling criteria (usually conflicting) CPU utilization, waiting time, response time, throughput, … Scheduling Algorithms FCFS, SJF, Priority, RR, Multilevel Queues, … Multiprocessor Scheduling Processor affinity vs. load balancing Evaluation of Algorithms Modeling, simulation, implementation Examples Win XP, Linux
Silberschatz, Galvin and Gagne ©2005 Operating System Concepts 5.36