CPU Scheduling:

1
CPU Scheduling

• Chapter 9

2
Goals of Scheduling

• Quick response time
• Fast throughput
• Processor efficiency

3
Type of Scheduling

• Long-term
• performed when new process is created
• Medium-term
• swapping
• Short-term
• which ready process to execute next

4
Long-Term Scheduling

• Determines which programs are admitted to the
  system for processing
• Controls the degree of multiprogramming
• More processes, smaller percentage of time each
  process is executed

5
Medium-Term Scheduling

• Main memory management
• Swapping
• Based on the need to manage multiprogramming

6
Short-Term Scheduling

• Determines which process gets the cpu.
• Only picks processes that are currently in
  memory (in the ready state /queue)
• Can not take a long time to pick.

7
Scheduling and Process State Transition
New
Long-term scheduling
Long-term scheduling
Exit
Running
Ready, suspend
Ready
Short-term scheduling
Medium-term scheduling
Blocked
Blocked, suspend
Medium-term scheduling
8
Short-Term Scheduling Criteria

• User-oriented
• Response Time
• Elapsed time between the submission of a request
  until there is output.
• System-oriented
• effective and efficient utilization of the
  processor

9
Short-Term Scheduling Criteria

• Performance-related
• measurable such as response time and throughput
• Not performance related
• predictability

10
Decision Mode Short term scheduler

• Nonpreemptive
• Once a process is in the running state, it will
  continue until it terminates or blocks itself for
  I/O
• Preemptive
• Currently running process may be interrupted and
  moved to the Ready state by the operating system
• Allows for better service since any one process
  cannot monopolize the processor for very long

11
Scheduler Goals

• Fairness
• - each process gets a fair share of CPU time
• Efficiency
• - keep the CPU busy
• Response Time
• - cant seem slow to users
• Turnaround
• - minimize the time batch users must wait for
  output
• Throughput
• - maximize the number of jobs processed and
  completed per unit time (more batch oriented)

12
Which goal?

• Cant do all of them - some goals contradict
• Old method
• - Batch system - one at a time - read the next
  set of cards or tape
• Multi-user systems required something better than
  sequential batch processing
• - Most unpredictable aspect is the amount of time
  waiting on I/O
• Run to completion (non-preemptive) vs. preemptive
• Old batch systems were run to completion which
  was simple and easy to implement
• Preemption enhances fairness and throughput, but
  leads to problems of synchronization

13
Scheduling Algorithms

• Common Scheduling Algorithms
• FCFS or FIFO
• Round-Robin
• Shortest Process Next or Shortest Job first
• Shortest Remaining Process Next
• Multi level feedback / Priority
• Highest Response Ratio Next
• Common measurements
• w time spent in system so far
• e time spent in execution
• s total time required by process
• Turnaround time total time in system (Tq)
• Normalized turnaround Tq/Ts

14
First-IN, First-OUT (FIFO) Scheduling

• Example Process Burst Time
• P1 24
• P2 3
• P3 3
• Suppose that the processes arrive in the order
  P1 , P2 , P3 The Gantt Chart for the schedule
  isAverage waiting time (0 24 27)/3
  17

15
First Come First Serve
16
First Come First Serve
17
FIFO Scheduling (Cont.)

• Suppose that the processes arrive in the order
• P2 , P3 , P1 .
• The Gantt chart for the schedule is
• Waiting time for P1 6 P2 0 P3 3
• Average waiting time (6 0 3)/3 3
• Much better than previous case.

P1
P3
P2
6
3
30
0
18
FIFO (First in first out)

• Very simple
• Non Preemptive
• Favors long process over short process
• Favors CPU bound over I/O bound
• Often used in combination with other algorithms

19
Alternating Sequence of CPU And I/O Bursts
20
Round Robin

• Very simple - Oldest fair algorithm
• No priorities - equal importance
• 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.
• Always have overhead of the process switch, so
  the quantum length can be important for
  efficiency
• q large ? FCFS
• q small ? q must be large with respect to context
  switch, otherwise overhead is too high.

21
Process set for our examples
Process Arrival Time Service Time
Assume process switch requires 3 units of
time. Assume 1 unit of time is required for
algorithm execution. Assume scheduling
algorithm itself is non Pre emptable.
22
Round Robin (q 5)
OS
0 3 10 18 26
34 38 46 53
61
Process
A 7 B 15 C 23 D
31 B 35 E 43 F 50 C 58
A - 4 B - 6 C - 13 D - 8 B - 1 E - 5 F - 4 C -
8 G - 12 H - 4 D - 3 C - 3 G - 7
61 69 76 82
88 94
OS
Process
G 66 H 73 D 79 C
85 G 93 G 96
23
Round Robin (q 5)
Process A/T Finish Ts
Turn
Tq/Ts /around
AVG 6.27
24
Round Robin

• Reduces bias in favor of long processes but....
• What about CPU bound processes?
• What are some solutions?

25
Shortest Process Next

• Predictability of longer processes is reduced
• If estimated time for process not correct, the
  operating system may abort it
• Possibility of starvation for longer processes
• Still penalizes short jobs that arrive after long
  ones begin

26
Shortest Process Next (Non- Pre-emptive)
OS
0 3 11 19 30
37 44 52 67
Process
A 7 B 16 D 27 F
34 H 41 E 49 G 64 C 80
A - 4 B - 6 C - 13 D - 8 C - 13 E - 5 F - 4 G -
12 H - 4
T 7
T 16
T 27
27
Shortest Process Next
Process A/T Finish Ts
Turn
Tq/Ts /around
AVG 3.33
28
Shortest Remaining Time

• Scheduling algorithm executes when either
• current process completes or
• new process arrives
• Scheduler chooses job with earliest expected
  completion time (including estimated process
  switch times)

29
Shortest Remaining Time
OS
0 3 6 11-12
14 22-23 26 28
33 40
Process
A 5 A 8 B 13
B 19 E 25 E 27 E 30 F
37
40 47 58 73
OS
Process
H 44 D 55 G 70 C 86
30
Shortest Remaining time (Pre-emptive)
Process A/T Finish Ts
Turn
Tq/Ts /around
AVG 3.70
31
Highest Response Ratio Next (HRRN)

• Scheduler only executes when a process completes
• Scheduler chooses the process with the largest
  value of

time spent waiting expected service time
expected service time
32
Highest Response Ratio Next (HRRN)
OS
0 3 10 19 35
42 50 57 68
Process
A 7 B 16 C 32 F
39 E 47 H 54 D 65 G 80
T 16 C 5 13/ 13 1.38 D 3 8/8 1
37
T 39 D 26 8/ 8 4.25 E 20 5/5
5 G 14 12 /12 2.16 H 12 4/4 4
T 47 D 34 8/ 8 5.25 G 22 12 /12
2.83 H 20 4/4 6
T 32 D 19 8/ 8 3.37 E 13 5/5
3.6 F 11 4/4 3.75 G 7 12 /12
1.58 H 5 4/4 2.25
T 54 D 41 8/ 8 6.12 G 29 12 /12 3.41
33
Highest Response Ratio Next (HRRN)
Process A/T Finish Ts
Turn
Tq/Ts /around
AVG 4.14
34
Feedback

• Penalize jobs that have been running longer
• Dont know remaining time process needs to execute

35
Multi-level Feedback
Arriving processes are inserted into level
1 Scheduling algorithm executes when job finishes
or quantum expires Picks the next process in the
lowest numbered queue
36
Multi-level Feedback
37
Multi-level Feedback
Process A/T Finish Ts
Turn
Tq/Ts /around
AVG 9.64
38
Priorities

• Scheduler chooses a process of higher priority
  over one of lower priority
• Have multiple ready queues to represent each
  level of priority
• Lower-priority may suffer starvation
• allow a process to change its priority based on
  its age or execution history

39
Multiple Priority Queue

• Some variation of this scheme is used in most of
  todays OS.
• A process can move between the various queues
  aging can be implemented this way.

40
Multiple Priority Queue

• 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

41
Recommended Practice
Process A/T Finish Ts
Turn
Tq/Ts /around
AVG 9.64
42
Recommended Practice

• Compete the the table using the following
  policies
• First Come First Served
• Round Robin (q 4)
• Shortest Process next
• Shortest Remaining time
• Highest Response Ration Next
• Multi- level feedback with unbounded number of
  levels starting at level 1 Processes in level i
  running for 3i units. IE q level i 3i)