Chapter 6:

Scheduling Algorithms
 Process Scheduling
 Process execution consists of a CPU burst followed by I/O
burst
 CPU burst distribution is of main concern

 Goal is to maximize CPU utilization

 Quickly switch processes onto CPU for
time sharing
 Select among available processes for
next execution on CPU
 Process Scheduling Diagram
 OS maintains scheduling queues of processes
 Job queue – set of all processes in the system
 Ready queue – set of all processes residing in main memory, ready
and waiting to execute
 Device queues – set of processes waiting for an I/O device
 Processes migrate among the various queues
Ready Queue And Various I/O Device Queues
 Schedulers
 Short-term scheduler (or CPU scheduler)
 Selects which process should be executed next and allocates CPU
 Sometimes the only scheduler in a system
 Invoked frequently (milliseconds)  (must be fast)
 Dispatcher module gives control of the CPU to the process selected by
the short-term scheduler
 Dispatch latency is the time it takes for the dispatcher to stop one
process and start another running

 Long-term scheduler (or job scheduler)

 Selects which processes should be brought into the ready queue
 Invoked infrequently (seconds, minutes)  (may be slow)
 Controls the degree of multiprogramming
 Strives for good process mix (I/O and. CPU bound processes)
 Physical Memory & Multiprogramming
 Memory is scarce resource
 Want to run many programs
 Programs need memory to run

 What happens when M(a) + M(b) + M(c) > physical memory?

 Medium Term Scheduler
 Medium-term scheduler can be added if degree of multiple
programming needs to decrease
 Remove process from memory, store on disk, bring back in from
disk to continue execution: known as job swapping
Histogram of CPU-burst Times
 Types of Scheduling
 Types of process scheduling

 Preemptive
Can be taken out from the processor after the time
slice gets expire

 Non-preemptive
Cannot be taken out unless a process completes its
execution
 Scheduling Criteria
 CPU utilization – keep the CPU as busy as possible

 Throughput – # of processes that complete their execution per unit time

 Turn around 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)
Scheduling Algorithm Optimization Criteria
 Maximize CPU utilization
 Maximize throughput
 Minimize turnaround time
 Minimize waiting time
 Minimize response time
 Process Scheduling
 Multiprogramming needs CPU scheduling
 Without any hardware support, what can the OS do to a running
process?

 System calls that triggers Scheduler

 Block – wait on some event/resource
 Network packet arrival (e.g. recv( ))
 Keyboard, mouse input (e.g. getchar( ))
 Disk activity completion (e.g. read( ))
 Yield – give up running for now
 Non-Preemptive Scheduler
 A non-preemptive scheduler invoked by explicit block/yield calls or
terminations
 Only method when there is no timer!

 The simplest form

 Scheduler:
 save current process state (into PCB)
 choose next process to run
 dispatch (load PCB and run)

 Used in Windows 3.1, Mac OS

 Timesharing Systems
 Timesharing systems support interactive use
 Each user feel he/she has the entire machine
 Optimize response time
 Based on time-slicing

 Without hardware support, can we do anything other than non-

preemptive scheduling?

 Timer interrupt
 Generated by the hardware
 Setting requires privilege
 Delivered to the OS
 Preemptive Scheduler
 Using interrupts for scheduling

 Basic Idea
 Before moving process to running, OS sets timer
 If process yields/blocks, clear timer, go to scheduler
 If timer expires, go to scheduler
 Recall Context Switching!

 Considerations: Timer granularity

 Finer timers = more responsive
 Coarser timers = more efficient
 OS as a Resource Manager:
Allocation vs. Scheduling
 Allocation (spatial)
 Who gets what? Given a set of requests for resources, which
processes should be given which resources for best utilization?

 Scheduling
 How long can they keep it? When more resources are requested than
can be granted, in what order can they be serviced?
 [Lecture 1] Separating Policy from
Mechanism

A fundamental design principle in Computer Science

 Mechanism – tool/implementation to achieve some effect

 Policy – decisions on what effect should be achieved
 Examples:
 All users treated equally
 All programs instances treated equally
 Preferred users treated better

Separation leads to flexibility

 Preemptive CPU Scheduling
 What is in it?
 Mechanism + Policy
 Mechanisms fairly simple
 Policy choices harder
 Challenges in Policy
 Flexibility – variability in job types
 Long vs. short
 Interactive vs. non-interactive
 I/O bound vs. compute-bound

 Issues?
 Short jobs shouldn’t suffer
 Users shouldn’t be annoyed
 Challenges in Policy (cont.)
 Fairness
 All users should get access to CPU
 Amount of CPU should be roughly even?

 Issues?
 Short-term vs. long-term fairness
 Goals and Assumptions
 Goals (Performance metric)
 Minimize turnaround time: average time to complete a job
 Maximize throughput: operations (jobs) per second
 Minimize overhead of context switches: large quanta
 Efficient utilization (CPU, memory, disk etc.)
 Short response time: type on a keyboard
 Small quanta
 Fairness (fair, no starvation, no deadlock)
 Goals often conflict
 Fairness vs. average turnaround time?

 Assumptions
 One process/program per user
 Programs are independent
 Scheduling Policies
 Is there an optimal scheduling policy?
 Even if we narrow down to one goal?

 But we don’t know about future

 Offline vs. online
 Scheduling Algorithms
 Non-preemptive
 First Come First Served
 Shortest Job First
 Priority Scheduling
 Preemptive
 Round Robin
 Shortest Remaining Time First
 Priority Scheduling
 Example: FCFS Scheduling Policy

Process Burst Time

P1 24
P2 3
P3 3
 Suppose that the processes arrive in the order: P , P , P
1 2 3
The Gantt Chart for the schedule is:

P1 P2 P3

0 24 27 30

 Waiting time for P = 0; P = 24; P = 27

1 2 3
 Average waiting time: (0 + 24 + 27)/3 = 17
 Example: FCFS Scheduling Policy
Suppose that the processes arrive in the order:
P2 , P3 , P1
 The Gantt chart for the schedule is:

P2 P3 P1

0 3 6 30

 Waiting time for P1 = 6; P2 = 0; P3 = 3

 Average waiting time: (6 + 0 + 3)/3 = 3
 Much better than previous case
 Convoy effect - short process behind long process
o Consider one CPU-bound and many I/O-bound processes
Class Exercise : FCFS with Arrival Time

Process Execution Time Arrival Time

P1 6 2
P2 3 5
P3 8 1
P4 3 0
 FCFS advantages and disadvantages
 Advantages
 Simple and easy to program

 Disadvantages
 Non-Preemptive scheduling algorithm so after the process has been
allocated to the CPU, it will never release the CPU until it finishes
executing or blocked.
 The Average Waiting Time is high.
 Short processes at the back wait for the long process at the front.
 Not an ideal technique for time-sharing systems.
Shortest-Job-First (SJF) Scheduling Policy
 Associate with each process the length of its next CPU burst
 Use these lengths to schedule the process with the shortest time
 Non-preemptive

 Advantage
 SJF is optimal – gives minimum average waiting time and
turnaround time for a given set of processes

 Disadvantage
 Can cause starvation if there is a steady supply of shorter processes
 Difficult to know the future (length of the next CPU request)
 Example: SJF Scheduling Policy
Process Burst Time
P1 6
P2 8
P3 7
P4 3

 SJF scheduling chart

P4 P1 P3 P2

0 3 9 16 24

 Average waiting time = (3 + 16 + 9 + 0) / 4 = 7

Class Exercise : SJF Scheduling Policy

Process Execution Time

P1 7
P2 8
P3 4
P4 3
Class Exercise : SJF Scheduling Policy

Process Execution Time

P1 7
P2 8
P3 4
P4 3

P4 P3 P1 P2

0 7 14 22
3

Average waiting time = (7 + 14 + 3 + 0) / 4 = 6

 SJF with Arrival time
Process Time Burst Time
P1 0.0 6
P2 2.0 8
P3 4.0 7
P4 7.0 3
 SJF with Arrival time
Process Time Burst Time
P1 0.0 6
P2 2.0 8
P3 4.0 7
P4 7.0 3

P1 P3 P4 P2

0 13 16 22
6
 SJF with Arrival time
Process Time Burst Time
P1 0.0 6
P2 2.0 8
P3 4.0 7
P4 7.0 3

P1 P3 P4 P2

0 13 16 22
6

Average waiting time = (0 + 14 + 2 + 6) / 4 = 5.5

 Determining Length of Next CPU Burst
 Can only estimate the length – should be similar to the previous one
 Then pick process with shortest predicted next CPU burst
 Can be done by using the length of previous CPU bursts, using
exponential averaging
1. t n  actual length of n th CPU burst
2.  n 1  predicted value for the next CPU burst
3.  , 0    1
4. Define :  n 1   t n  1    n .
 Commonly, α set to ½
  =0
n+1 = n (recent history does not count)
  =1
 n+1 =  tn (only the actual last CPU burst counts)
Prediction of the Length of the Next CPU Burst
 SRTF Scheduling Policy
 Shortest Remaining Time First scheduling
 The preemptive version of SJF.
 Calculate the remaining execution time for each process
 Schedule the process with the minimum remaining time

 Advantage
 Minimal average waiting and turnaround time for a given set of
processes

 Disadvantage
 Like SJF, can cause starvation for longer processes
 Example: SRTF Scheduling Policy
 Now we add the concepts of varying arrival times and preemption to the
analysis

Process Arrival Time Burst Time

P1 0 8
P2 1 4
P3 2 9
P4 3 5
 Preemptive SJF Gantt Chart

P1 P2 P4 P1 P3

0 1 5 10 17 26

 Average waiting time = [(10-1)+(1-1)+(17-2)+5-3)]/4 = 26/4 = 6.5 msec

Class Exercise: SRTF Scheduling Policy

Process Arrival Time Execution Time

P1 0 7
P2 1 3
P3 2 8
P4 3 4
Class Exercise: SRTF Scheduling Policy

Process Arrival Time Execution Time

P1 0 7
P2 1 3
P3 2 8
P4 3 4

P1 P2 P4 P1 P3

8 14 22
0 1 4

P1 = 8 – 1 = 7
P2 = 1 – 1 = 0
P3 = 14 – 2 = 12
P4 = 4 – 3 = 1
Average waiting time = (7 + 0 + 12 + 1) / 4 = 5
 Priority Scheduling
 Assign each process a priority (integer) number (smallest integer  highest
priority)
 Run the process with highest priority in the ready queue first
 Preemptive
 Non-preemptive
 Use FIFO for processes with equal priority
 SJF is priority scheduling where priority is the inverse of predicted next
CPU burst time

 Advantage
 Flexibility: Not all processes are born equal

 Problem  Starvation – low priority processes may never execute

 Solution  Aging – adjust priority dynamically (increase priority over time)
 Example: Priority Scheduling
Process Burst Time Priority
P1 10 3
P2 1 1
P3 2 4
P4 1 5
P5 5 2

 Priority scheduling Gantt Chart

P2 P5 P1 P3 P4

0 1 6 16 18 19

 Average waiting time = 8.2 msec

 Priority (preemptive)
 A number is assigned to each process that indicates its priority
level.

 If a newer process arrives,

 having higher priority than the currently running process,
 then the currently running process is preempted.
 Important Points
 The waiting time for the process having the highest priority
 will always be zero in preemptive mode.
 may not be zero in non-preemptive mode.
 Preemptive and non-preemptive mode behaves exactly same,
if
 The arrival time of all the processes is same
 All the processes become available
 Exercise

Arrival Burst
Process Id Priority
time time

P1 0 4 4

P2 1 3 3

P3 2 1 2

P4 3 5 1

P5 4 2 1
 Exercise

Arrival Burst
Process Id Priority
time time

P1 0 4 4

P2 1 3 3

P3 2 1 2

P4 3 5 1

P5 4 2 1
 Exercise

Arrival Burst
Process Id Priority
time time

P1 0 4 4

P2 1 3 3

P3 2 1 2

P4 3 5 1

P5 4 2 1

P1=11 P2=8 P3=0 P4=0 P5=4

Avg. waiting time=4.6
 Round Robin (RR) Scheduling Policy

 Each process runs a time slice or quantum q.

 For n processes in the ready queue and time quantum is q, each process gets
1/n of the CPU time in chunks of at most q time units. Each process waits
no longer than (n-1)q time units until its next time quantum.
 Timer interrupts every quantum to schedule next process
 Performance
 q large  FIFO
 q small  q must be large with respect to context switch, otherwise
overhead is too high
 Example: RR Scheduling Policy
Process Burst Time
P1 24
P2 3
P3 3
Time slice = 4
 The Gantt chart is:

P1 P2 P3 P1 P1 P1 P1 P1

0 4 7 10 14 18 22 26 30

 Typically, higher average turnaround time than SJF, but better response
time
 q should be large compared to context switch time
 q usually 10ms to 100ms, context switch < 10 usec
 RR
 Advantages
 best performance in terms of average response time.
 best suited for
 time sharing system,
 client server architecture and
 interactive system.
 Disadvantages
 leads to starvation for processes with larger burst time
 as they have to repeat the cycle many times.
 Its performance heavily depends on time quantum.
 Important Notes
 With decreasing value of time quantum,
 Number of context switch increases
 Response time decreases
 Chances of starvation decreases
 When time quantum tends to infinity, Round Robin Scheduling becomes
FCFS Scheduling.
 The value of time quantum should be
 neither too big nor too small.
Turnaround Time Varies With The Time Quantum

80% of CPU bursts

should be shorter
than q.
 Exercise
Process Id Arrival time Burst time

P1 0 5

P2 1 3

P3 2 1

P4 3 2

P5 4 3

Time slice = 2
sec
 Exercise
Process Id Arrival time Burst time

P1 0 5

P2 1 3

P3 2 1

P4 3 2

Time slice = 2 P5 4 3
sec

Average waiting time = (8 + 8 + 2 + 4 + 7) / 5 = 29 / 5 = 5.8

 Multilevel Queue Scheduling
 When processes can be readily categorized
 then multiple separate queues can be established
 each implementing whatever scheduling algorithm is most
appropriate for that type of job, and/or with different
parametric adjustments.

 Scheduling must also be done between queues

 Multilevel Queue Scheduling
 Two common options are
 strict priority
no job in a lower priority queue runs until all higher
priority queues are empty
 round-robin
each queue gets a time slice in turn, possibly of different
sizes
 Note that under this algorithm jobs cannot switch from queue
to queue
 Once they are assigned a queue, that is their queue until
they finish.
MULTILEVEL FEEDBACK-
QUEUE SCHEDULING
 Multilevel Feedback-Queue Scheduling
 Similar to the multilevel queue scheduling
 except jobs may be moved from
 one queue to another for a variety of reasons

 Most flexible
 because it can be tuned for any situation.

 Most complex to implement

 because of all the adjustable parameters.
 Evaluation of Scheduling Policies
 How to select CPU-scheduling policy for an OS?
 Determine criteria, then evaluate algorithms
 Deterministic modeling
 Type of analytic evaluation
 Takes a particular predetermined workload and defines the performance
of each algorithm for that workload
 Consider 5 processes arriving at time 0:
 FCFS, SJF, and RR
 Deterministic Evaluation
 For each algorithm, calculate minimum average waiting time
 Simple and fast, but requires exact numbers for input, applies only to
those inputs
 FCFS is 28ms:

 Non-preemptive SJF is 13ms:

 RR is 23ms: